METHODS FOR IDENTIFICATION OF NOVEL PROTEIN DRUG TARGETS AND BIOMARKERS UTILIZING FUNCTIONAL NETWORKS
The process of System Reconstruction is used to integrate sequence data, clinical data, experimental data, and literature into functional models of disease pathways. System Reconstruction models serve as informational skeletons for integrating various types of high-throughput data. The present invention provides the first metabolic reconstruction study of a eukaryotic organism based solely on expressed sequence tag (EST) data. System Reconstruction also provides a method for the identification of novel therapeutic targets and biomarkers using network analysis. The initial seed networks are built from the lists of novel targets for diseases with the high-throughput experimental data being superimposed on the seed networks to identify specific targets.
Latest GENEGO, INC. Patents:
- SYSTEM AND METHOD FOR PREDICTION OF DRUG METABOLISM, TOXICITY, MODE OF ACTION, AND SIDE EFFECTS OF NOVEL SMALL MOLECULE COMPOUNDS
- System reconstruction: integrative analysis of biological data
- Methods for identifying compounds for treating disease states
- Methods for identification of novel protein drug targets and biomarkers utilizing functional networks
This application is a continuation of U.S. application Ser. No. 13/210,302, filed on Aug. 15, 2011, now pending, which is a continuation of U.S. application Ser. No. 11/499,437, filed on Aug. 4, 2006, now U.S. Pat. No. 8,000,949, issued Aug. 16, 2011, which is a continuation-in-part of U.S. application Ser. No. 10/518,103, filed on Oct. 14, 2005, now U.S. Pat. No. 8,000,948, issued Aug. 16, 2011, which is the National Stage Entry of PCT International Application No. PCT/US03/19325, filed on Jun. 18, 2003, published as WO 03107545, which claims the benefit of priority of U.S. Provisional Patent Application No. 60/389,474, filed on Jun. 18, 2002, and U.S. application Ser. No. 10/174,762, filed on Jun. 18, 2002, now U.S. Pat. No. 8,010,331, issued Aug. 30, 2011, which claims the benefit of priority of U.S. Provisional Patent Application No. 60/299,040, filed on Jun. 18, 2001. All of these applications are herein incorporated by reference in their entirety.
REFERENCE TO SEQUENCE LISTING SUBMITTED VIA EFS-WEBThe entire content of the following electronic submission of the sequence listing via the USPTO EFS-WEB server, as authorized and set forth in MPEP §1730 II.B.2(a), is incorporated herein by reference in its entirety for all purposes. The sequence listing is identified on the electronically filed text file as follows:
The present invention relates to bioinformatics technologies. More specifically, the present invention relates to the technology of System Reconstruction. The present invention further relates to methods for elucidating metabolic pathways for the identification of novel therapeutic targets and biomarkers using network analysis. The initial seed networks are built from the lists of novel targets for diseases with the high-throughput experimental data being superimposed on the seed networks to identify specific targets.
BACKGROUND ARTThe past few years have seen dramatic advances in genomics and other areas of high-throughput biology. The fruits of these accelerated technologies culminated in last-years publication of the human genome. The availability of the DNA sequence of the human genome promises to alleviate much of human suffering from life-threatening diseases. Knowledge of an entire genome may lead to the discovery of new drug targets. Access to the DNA sequence of an individual promises to reduce drug side effects and to allow tailoring medicine to the individual's genetic makeup. Both government agencies and drug companies have invested heavily in these technologies. In return, they expected to vastly reduce the cost and time of drug development, a process costing on average over $500 million in the 1990s and usually spanning over a decade from the initial discovery of drug targets and leads, through validation, optimization, and finally clinical trials.
Currently, these expectations are far from reality because human biology is complex, and there has been no systematic approach to capture this biological complexity. A new field of computational biology has been forged to make sense out of the inordinate amount of genomics data including DNA sequence data, gene expression data, proteomics, metabolomics, and cellomic data. It is believed by many in the industry that the integration of these data alone would quickly lead to the correlation of phenotype (clinical manifestations) with genotype (variations in gene sequence). That goal is still far off, however, as the majority of these data are examined out of context. The basis of a disease cannot be understood without understanding, for example, the alternative splicing forms of the related genes, the proteins for which they code, the complex networks of protein interactions involved, the multiple levels of gene regulation and expression, the correlations between healthy and diseased tissue, the significance of clinical data, and the like. The complexity of human biology requires a systemic understanding of genomic data rather than a shotgun understanding. As a result, the field of systems biology arose and is rapidly becoming a leading approach to understanding human biology.
Recent progress in sequencing technology has generated a vast amount of genomic data. According to the GOLD database, there are more than 300 genomic projects currently completed or under development (wit.integratedgenomics.com/GOLD/). Seventy-nine complete or partially complete genomes are available through the public ERGO system (igweb.integratedgenomics.com/lGwit/). In order to handle this wealth of information, several powerful bioinformatics systems have been developed. The WIT Project was instituted to develop a framework for the comparative analysis of genomic sequence data, focusing largely on the development of metabolic models for sequenced organisms. The analysis of the genomes involves several distinct, but complementary efforts. The first is a determination of open reading frames (ORFs). The second, often called annotation, is the assignment of functions to genes. The third is the creation of functional models for metabolic and regulatory networks of the sequenced genomes, referred to as reconstruction.
Metabolic reconstruction for bacterial and archaeobacterial genomes has been carried out. In contrast, metabolic reconstruction for eukaryotic organisms remains a much more complicated problem. Despite significant progress in genome sequencing, the annotation of eukaryotic genomes remains a complicated problem. Even finding the ORFs, a key component of gene identification, is still a very difficult task. A comprehensive understanding of the complicated structure of eukaryotic genomes will require the integration of sequencing information with genetic, biochemical, structural, and evolutionary data. It will require developing new bioinformatics tools and discovering new algorithms, and, most likely, it will take years of research in both dry and wet labs.
Traditionally, it has not been considered feasible to study metabolism based on expressed sequence tag (EST) data. Such an approach, however, would be very useful for comparative analyses of complex eukaryotic genomes. First, generation of a complete set of ESTs is at least an order of magnitude less expensive than whole genome sequencing. Second, there is a great deal of processed EST data freely available to the scientific community. Currently, there are only a few complete eukaryotic genomes available to the public, but there are sufficient EST data for several dozens of species. Third, and most important, ESTs represent genes that are expressed at specific times in specific tissues. In the present invention, expressed sequence tag data, rather than genomic sequences, were used to reconstruct various aspects of human metabolism.
Several databases exist for collecting EST sequence and expression patterns for eukaryotic genes (for example Unigene EST, dbEST, STACK, SAGE, DOTS, trEST, XREFdb, in addition to a number of tissue-specific databases, such as PEDB). A significant amount of human EST data has already been carefully analyzed, classified, annotated, and mapped to chromosomes. Currently, there are over 1,000,000 human ESTs available in public databases representing 50-90% of all human genes. It is generally believed, however, that EST sequences are inferior to genomic DNA sequences in terms of their quality and degree of representativeness.
Additionally, numerous public and commercial efforts that have focused on characterizing various aspects of general biochemistry and metabolism. Some of these databases include KEGG, BRENDA, SWISS-PROT, EcoCyc, and EMP/MPW. None of these databases, however, focus specifically on humans, or on a single species.
The technology known as Metabolic Reconstruction was developed by Dr. Evgeni Selkov and co-workers at the Argonne National laboratory. Metabolic Reconstruction was developed to study an organism's metabolism by using its genome sequence. A reconstruction of the metabolism of Methanococcus jannaschii from sequence data can be found in Gene, 197, GC11-26.
Cellular life can be represented and studied as the interactome the dynamic network of biochemical reactions and signaling interactions between active proteins. Systemic networks analysis is optimal for integration and functional interpretation of high-throughput experimental data which are abundant in drug discovery yet poorly understood. Composition and topology of complex networks are closely associated with vital cellular functions, which have important implications for life science research. Network theory advances has, in recent years, quickly advanced; and reliable databases of protein interactions for human and model organisms and comprehensive analytical tools have become available. In this application, we present a specific application of networks analysis: identification of novel drug targets by reverse engineering the networks which connect the existing targets for specific disease, followed by superposition of experimental molecular data such as microarray gene expression, proteomics and metabolomics.
Over the last several years known as the post-genomics era, we have seen a paradigm shift in life science research due to the unprecedented scale-up of several laboratory techniques such as automated DNA sequencing, global gene expression measurements, and proteomics and metabolomics techniques. The high throughput (HT) data collectively referred to as OMICs are ubiquitous throughout the drug discovery pipeline from target identification and validation to the development and testing of drug candidates to clinical trials. However, OMICs data is poorly utilized due to the lack of the adequate methods for interpretation in the context of disease and biological function. Although bioinformatics has developed robust statistical solutions for evaluation of the significance and clustering the data points, statistics alone do not explain the underlying biology.
The complexity of human biology requires a system-wide approach to data analysis, which can be defined as the integration of OMICs data using computational methods. The field states that the identification of the parts list of all the genes and proteins is insufficient to understand the whole. Rather, it is the assembly of these parts (the general schema, the modules and elements) and the dynamics of changes in response to stimuli that is truly the key to understanding life, form and function. The assembly of cellular machinery is to be most properly presented as the interactome, the network of interconnected signaling, regulatory and biochemical networks with proteins as the nodes and physical protein-protein interactions as edges. Across many fields of science, technology and social life, the topology and dynamics of complex networks are studied by graph theory. The information about protein interactions has being collected from the vast published experimental data, which is annotated and assembled in the interactions databases. The network data analysis that are now commercially available are robust enough for simultaneous processing of dozens of multi-thousand featured strong data files such as whole-genome expression microarrays. Just recently, researchers in systems biology announced the interpretation of experimental OMICs datasets in the context of accumulated knowledge on human functional networks as the first step in studying complex systems. With this development, the building of the basic framework of databases and logistics can be considered completed. Networks-centered data analysis is now well underway at the major pharmaceutical companies.
SUMMARY OF THE INVENTIONThe process of the present invention, referred to as System Reconstruction, integrates data on organism- and tissue-specific biochemical pathways, genome sequences, conditional gene expression, and genetic polymorphisms with clinical manifestations of diseases and other clinical traits. As a result, a network of interconnected functional pathways (a Functional or System Model) is constructed in which elements are linked to appropriate molecular data (ORFs, ESTs, SNPs, etc.) and annotated with relevant clinical information.
Generally, the first step in creating a System Reconstruction model is the determination of a network of relevant biochemical pathways, specific for certain human tissues at certain developmental stages (Metabolic Reconstruction). Next, the collection of pathways is extended by computational reconstruction of relevant metabolic networks. Third, the expression data is integrated into the resulting metabolic map to generate a snapshot for any specific cell, organ, or tissue. Comparison of such snapshots constructed for the same tissue in normal and disease states (or in different developmental stages), provides valuable information about regulatory mechanisms of the disease or of development. Finally, the System Reconstruction model is completed by integrating the developmental pathways and mapping them onto the metabolic network. This step verifies the regulatory pathways and completes the functional overview of the network.
In one aspect, the present invention ascertains necessary functions involved in a particular metabolic pathway.
In another aspect, the present invention provides a visual overview of I expressed genes associated with a particular pathway specific for normal and abnormal human tissues.
In another aspect, the present invention provides a method for I determining and identifying the ORFs involved in those pathways.
In another aspect, the present invention provides a method for comparing System Reconstructions made for normal and diseased organs or tissues, thus providing important information about possible regulatory mechanisms and potential drug targets. In another aspect, the present invention provides a method for comparing the reconstructions made for the same tissue at different developmental stages, thus providing information about the developmental timing of gene expression and revealing possible targets for gene therapy.
In another aspect, the present invention provides a method for I mapping single nucleotide polymorphism (SNP) sites to corresponding metabolic genes and/or predicted ORFs, thus providing physiological insights into associations of SNPs with unknown phenotypes.
The present invention relates to a method for determining necessary functions involved in a particular metabolic pathway. In one aspect, the present invention provides a visual overview of expressed genes associated with a particular pathway specific for normal and abnormal human tissues. The present invention can also provide a method for determining and identifying the ORFs involved in those pathways. The present invention further provides a method for comparing System Reconstructions made for normal and diseased organs or tissues, thus providing important information about possible regulatory mechanisms and potential drug targets.
In another aspect, the present invention provides a method for comparing the reconstructions made for the same tissues at different developmental stages, thus providing information about the developmental timing of gene expression and revealing possible targets for gene therapy.
In another aspect, the present invention provides a method for mapping single nucleotide polymorphism (SNP) sites to corresponding metabolic genes and/or predicted ORFs, thus providing physiological insights into associations of SNPs with unknown phenotypes.
The present invention also relates to the determination of complicated cellular networks using abundant gene expression data (such as EST and micro-array data) as well as genomic sequence data; the identification of relationships between different human genes, pathways and parts of metabolism the identification and grouping according to function of over- and under-expressed genes specific for given tissue or condition; the generation of interactive, integrated functional outlines for all parts of human metabolism.
Identification of novel therapeutic targets using network analysis. The initial seed networks are built from the lists of novel targets for diseases. The high-throughput experimental data is superimposed on the seed networks to identify specific targets.
The accompanying drawings are not intended to be drawn to scale. For purposes of clarity, not every component may be labeled in every drawing. In the drawings,
A bioinformatics approach called System Reconstruction is used to integrate clinical information with high-throughput molecular data. In the core of this approach, a collection of human tissue-specific and condition-specific biochemical pathways are linked by common intermediates into maps or models. These models serve as a framework to integrate complementary types of high-throughput data and to establish mechanisms underlying clinical manifestations of diseases.
The present invention creates a system that allows building human-specific system-level models of biochemistry. In summary, information regarding human-specific pathways is collected. The pathways are linked to functional information, disease manifestations, and high-throughput data. Finally, pathways are connected to each other and linked to relevant; information to form a functional model. These models can be used, for example, as skeletons for further integration of high-throughput data, for deciphering mechanisms of diseases, for predicting drug metabolism and toxicity, and the like. System Reconstruction is a complex multi-step process that involves assembling a collection of human-specific pathways and results in fully annotated interactive maps of specific metabolic systems (see
The process of System Reconstruction generally starts with the creation of a collection of metabolic pathways. Pathways that are human specific and in the form in which they occur in humans are included. Building such a collection is achieved through a multi-level annotation process. Starting with a collection of identified metabolic pathways from mammals and non-mammals, the pathways are divided into categories based on relevance. For example, pathways are ranked according to the probability of their relevance in human metabolism. The most relevant pathways include multi-step mammalian pathways in which all of the reactions are catalyzed by identified human enzymes or at least enzymes that have ORE candidates in the human genome. Less relevant pathways include, for example pathways in which the necessary enzymes have not been identified in humans, and single step pathways. Information such as clinical data and scientific literature is reviewed to confirm which pathways are, in fact, present in humans.
In order to organize the information collected in the process of reconstruction, a relational database has been developed using Oracle RDBMS. Unlike many biomedical databases which are centered around a certain theme (e.g. sequences, proteins, biochemical reactions, etc.), the database developed in the present invention is a polythematic database that is built around several central data entities and relations among them. These entities are enzymes; compounds; reactions; pathways; genes; and diseases. This core architecture provides multiple linking portals for including other often heterogeneous data such as gene expression, protein interactions, metabolite profiles, etc. Once linked, these data become a part of a large system-level picture.
Currently, the database contains about 3300 pathways described in various species of mammals and about 2060 non-mammalian pathways. Of the mammalian pathways about 920 are multi-step pathways and the rest are single-step pathways. The pathways are divided into several categories according to the probability of their relevance to human metabolism. The most relevant category includes multi-step mammalian pathways for which all reactions are catalyzed by either identified human enzymes or enzymes that have ORF candidates in the human genome (about 710 pathways). The next category includes multi-step mammalian pathways that have human enzymes at the beginning and at the end of the pathway (about 40 pathways). In the next category, there are mammalian and non-mammalian multi-step pathways that contain human enzymes in the middle of the pathway (about 800 pathways). Finally, there are pathways with no identified human enzymes (about 1500 pathways).
In addition to these categories, there is a collection of single step reactions that can be catalyzed by human enzymes (about 2300 pathways) or by mammalian enzymes (over 5000 pathways). It should be noted, however, that not every such reaction, which can be catalyzed by a human enzyme, is in fact a functional human pathway. Many enzymes possess a broad spectrum of specificity in vitro, while in vivo there are many additional constraints that limit their functionality such as, e.g., compartmentalization, absence of precursors, and kinetic competition.
The process of ranking, as described above, creates a working collection of pathways that are then annotated. The initial collection of pathways may contain many pathways that are similar to human pathways but still have essential differences. Some of the differences may be in cofactors or sub-cellular localization of enzymes and metabolites. Also, human versions of pathways may be truncated or contain additional steps when compared to pathways from other species. Since many enzymes show a range of specificity, they may substitute for each other in similar pathways from different species. Therefore, during the annotation process, the available literature for every pathway is reviewed to determine the human specific form of the pathway. Pathways from the two most relevant categories are usually easy to verify through biomedical literature and generally require few, if any, modifications. The third category of pathways, as well as single step reactions with human enzymes, generally require a thorough literature search to be confirmed or rejected as human-specific pathways and usually undergo substantial changes. Finally, pathways with no human enzymes are left until the later stages when metabolic maps are built. At that point, some of those pathways are selected as candidate human pathways if they fit well into gaps in the map that cannot be easily filled by pathways from higher-ranking categories.
In addition to creating a collection of human specific pathways, the process of annotation yields important functional data about each pathway and its elements. In order to structure this information, a pathway is described as a hierarchy of biochemical units. These units comprise the pathway itself, individual steps that make up the pathway, chemical compounds, reactions, and enzymatic functions that are involved in each step. Enzymatic functions are related, in turn, to molecular species-specific proteins and genes.
In a process called structured annotation, explicit and implicit links are established between particular biochemical units and specific categories and instances in other data fields, discussed in greater detail below. Practically, this is achieved by filling in annotation tables associated with each biochemical unit. Examples of fields in these tables include: organ and tissue localization of the unit; intracellular localization and/or compartmentalization; existence and sub-cellular localization of the unit in other organisms; connection of the unit with inherited and common diseases and other functional disorders; type of relationship between the unit and a disease (e.g., cause, manifestation, etc.); references on the information source; and the like. The individual data fields can be linked in numerous ways including finding compounds, enzymes, reactions, and pathways that are directly linked in a particular unit; automatically interconnecting pathways and reactions into networks based on shared intermediates or other links; establishing constraints on pathway interactions based on sub-cellular localization of their components; finding pathways, reactions compounds, and enzymes related to a disease, its causes or manifestations, and interconnecting such elements into a disease network; finding diseases related by common pathways, reactions, or compounds; and finding alternative pathways for degradation or biosynthesis of specific compounds, to circumvent certain enzymes.
Thus, in the architecture of the database, functions can have a role as space-holders (
As illustrated in
Preferably, the database architecture accounts for various complexities of metabolism. For example, most enzymes can catalyze a range of reactions, and many reactions can be catalyzed by more than one enzyme. This multiplicity is preferably represented in a System Reconstruction database. As another example, there is usually more than one gene that corresponds to an enzyme or enzymatic function. There are currently about 2000 human genes assigned to enzymes corresponding to about 800 EC numbers. This type of multiplicity can also be represented in a database according to the present invention.
The next step is the building of functional models of specific categories of human metabolism, diseases, and other system-level reconstructions. Two important steps are (1) selecting a subset of the relevant pathways, and (2) linking them into metabolic networks. The selection of pathways is done by a set of “SELECT . . . FROM . . . WHERE . . . ” type queries, relying on the information collected in the structured annotation tables discussed above. The information on links among pathways is implicitly contained in the database. For example, whenever two pathway records share a common intermediate, or when an intermediate in one pathway occurs as a regulatory factor in a record for an enzyme from another pathway, a link is generated between the two pathways. Further computations are facilitated when such links are translated into explicit relations among pathways. To this end, stoichiometric matrices that represent the participation of compounds in the reactions are assembled. Using these matrices, it is possible to find links among reactions and, since reactions are already related to pathways in the database, a network of interconnected pathways can be generated.
At this stage, such networks are considered crude skeletons and are likely to contain substantial gaps as well as many nonfunctional links among pathways. A careful review and modification is undertaken to develop approved functional models. To fill in gaps, a set of candidate pathways is chosen from pathways of closely related organisms as well as from hypothetical pathways, and constructed by formally linking reactions. Then genomic DNA and ESTs are used as additional evidence to validate the proposed pathways.
It should be noted that the quality of stand-alone eukaryotic ESTs is often not sufficient for unambiguous functional assignments. However, if functional assignments are done with additional constraints imposed by a skeletal functional model, the ambiguity generally can be eliminated. In other words, an initial functional model provides insight into the work plan of a specific biochemical system, thereby allowing other data to be analyzed within the context of this work plan.
At this stage, sets of enzymatic functions that participate in the hypothesized pathways are identified and a determination is made as to which ones can be verified by sequence and expression data. Those that are supported by this evidence are added to the model as proposed pathways. It is also possible to consider other types of high-throughput data including metabolic profiles, two-hybrid assays, and other types of data to further validate these pathways. The proposed pathways can become primary targets for further experimental research. For the resulting network, the information on diseases associated with pathways, enzymes, and compounds is extracted from structured annotations and explicitly related to corresponding elements. The reconstruction is represented as an interactive map from which other information can be accessed, as described below.
The database developed according to the present invention can address various problems that often result from the traditional view of metabolism. The database can provide a representation of a wide spectrum of enzyme activity. Current enzyme nomenclature is built on the assumption that there is a single enzyme for each enzymatic reaction. This assumption is not always true in practice. Many enzymes can catalyze a range of reactions, and many reactions can be catalyzed by more than one enzyme. The database developed according to the present invention can represent this multiplicity by introducing many-to-many relations between enzymes and reactions.
The database can reflect the relationships between enzymatic function and molecular species. The term “enzyme” is somewhat ambiguous. While some biologists apply it to a particular protein—a molecule of certain chemical composition (or a complex of a few proteins)—, others refer to the function itself—an ability to catalyze a certain type of reaction. In the data model according to the present invention, this ambiguity is avoided by establishing several entities that are related to the term “enzyme”. One such entity is enzymatic function which is an ability to catalyze a certain reaction or class of reactions. Enzyme nomenclature and EC numbers are used to classify functions. Relating to any given function, there are specific molecular entries, such as proteins and genes. This system avoids the ambiguity that can occur when a single protein may possess a spectrum of catalytic activities, or when there may be more than one protein capable of catalyzing a certain reaction. In addition to avoiding ambiguity, such a data model is extremely useful in the process of functional annotation. For example, a disease that is linked to an enzymatic deficiency could have many potential causes, such as a mutation in the gene coding for the enzyme, problems at the gene expression level, or protein mis-folding, to name a few. This expanded data model allows the association of a clinical trait with the appropriate specific data entity.
The database also addresses the compartmentalization and localization of enzymes and metabolites. In living cells, reactions take place in certain compartments and intracellular localizations. This is one of the major mechanisms that cells use to regulate intracellular processes. Many enzymes have a fairly broad spectrum of substrates. Specificity is often determined by co-localization of an enzyme and one of its substrates. In some cases, incorrect protein localization is implicated in a disease. This type of information is included in the database by developing a representation of cellular anatomy. Preferably, compartments and organelles found in different cell types and their mutual arrangement are reflected in the database. Spatial organization of metabolic processes is represented by establishing relationships between anatomical data and data on pathways, reactions, enzymes, and compounds.
The technology of the present invention was used to build the System Reconstruction of amino acid metabolism in human, a portion of which is illustrated in
A user can also retrieve information about the involvement of a particular pathway, reaction, or enzyme for a specific disease. Preferably, structured annotations are accessible for the elements of the network (e.g., for pathways, reactions, enzymes, and the like) that specify whether the element is the cause of the disease or a manifestation of the disease (part of the disease fingerprint). In addition, a user is able to cross-link among the biochemical fingerprints of different diseases. The information is accessible by clicking on from the corresponding objects on the graphical map.
Pathways are interconnected into a network by shared metabolites. By clicking the mouse on a pathway or a component of a pathway, a user can access the pathway page (
One feature of the reconstruction is the incorporation of human diseases. By activating a link to diseases, a user can see lists of diseases associated with the pathway (
One aspect of the System Reconstruction technology of the present invention is that it uses organism specific pathways to build maps. This allows the imposition of a condition of self-consistency on the resulting networks. This means that each metabolite should either be essential for the organism (e.g., consumed through food) or there should be a pathway that produces it. In other words, if there is a gap between two nonessential compounds, this implies a lack of knowledge and serves to direct further research. This allows the prediction of the existence of an enzyme function in an organism even if organism-specific genes or proteins have not been identified. For example, when there is a clear gap between two metabolites in the reconstruction that cannot be filled in by any of the described enzymes, it is predicted that there is at least one undescribed enzyme that bridges this gap. In the present reconstruction of amino acid metabolism in humans, several human enzymes were identified that had not been previously identified in the human genome. These enzymes, including amino carboxymuconate-semialdehyde decarboxylase (EC 4.1.1.45) and imidazolone-5-propionate hydrolase (EC 3.5.2.7), were identified because their functions were required by the logic of the metabolic map. Consequently, human genes for these enzymes were proposed through thorough similarity searches of the human genome and by studying human ESTs.
The self-consistency condition also helps eliminate pathways that might be incorrectly assigned merely on the basis of human enzymes having been identified. One example can be illustrated with phenylalanine biosynthesis. It is well known that humans cannot synthesize this essential amino acid. However, there is a human enzyme, aspartate transaminase (EC 2.6.1.1), that could potentially synthesize phenylalanine from phenyl pyruvate. Simply superimposing the human enzyme onto a general metabolic map would lead to the incorrect conclusion that there is a human pathway for phenylalanine biosynthesis. In contrast, the self-consistent reconstruction of the present invention shows that the absence of phenyl pyruvate, the substrate for aspartate transaminase, makes biosynthesis of phenylalanine improbable in humans.
Examples 1 through 4 illustrate pathways in which chitinase is involved. These pathways have been elucidated through the use of the System Reconstruction technology.
Another important feature of the System Reconstruction technology is its potential to predict novel human pathways that have not yet been discovered. Indeed, only a fraction of human functional pathways have been described experimentally. There are still many unknown regulatory, signaling, and even metabolic pathways. At present, there are about 2,000 identified human enzymes. According to both Celera and the Public Human Genome Project Consortium, about 10% of human genes are involved in metabolism. Therefore, humans may have 3,000-4,000 metabolic enzymes in total. Thus, approximately half of the human metabolic enzymes may still need to be identified. System Reconstruction technology enables the proposal of many of these undiscovered human enzymes in the course of creating functional tissue-specific maps. The architecture of the map, including identified pathways, compounds that have been synthesized by these pathways, as well as additional evidence from literature and biological high-throughput data can point to enzymatic functions that are required for the self-consistency of the model, thus identifying undiscovered enzymes.
In one preferred embodiment of the present invention, the subject of System Reconstruction is human metabolism. System Reconstruction can be used to study diverse processes including, but not limited to, amino acid metabolism; carbohydrate metabolism; lipid metabolism; hormones; DNA, RNA, and nucleotide metabolism (see,
Generally, a formal network would contain reactions that are linked by shared metabolites. In System Reconstruction, pathways are also confirmed through a process of annotation. System Reconstruction allows building of both formal networks, which may contain putative pathways, as well as reconstructed pathways that have been confirmed through a process of annotation.
One example of a database architecture according to the present invention is illustrated in
Orgs, and OrgRels includes information about the organism and its taxonomic classification;
Locs includes information about the sub-cellular localization;
Tiss includes information about the tissues and organs in which the entity is present;
Chems, and ChemNames includes information about chemical compounds, their names, and synonyms;
Compas includes information about unique combinations such as a chemical and its sub-cellular localization (for example, glucose in cytoplasm);
Reacts includes information about reactions;
Rcomps includes information about links between the Reacts and Compas categories (for example, a chemical formula or reaction and its sub-cellular localization);
ReactOrgs includes information about organisms and tissues in which a reaction occurs;
Functions, and FuncNames includes information about enzymes, their EC numbers, their names, and their synonyms;
FuncOrgs includes information about organisms, tissues in which an enzyme is present as well as information about sub-cellular localizations;
ReactEC includes information about links between enzymes and reactions, showing which enzyme(s) catalyze a given reaction;
Pathways includes information about pathways, or sequences of several reactions;
PwReacts includes information about the reaction composition of a pathway;
Prots includes information about proteins, including the name and function of the protein;
ProtEC includes information about which human proteins correspond to a given function (EC number);
SwissProt, and ProtMIMs provide links to external protein databases;
Genes, and GeneNames include information about genes, their names, and their functions;
GeneProts, and GeneEC includes information about links between genes, proteins and EC numbers;
GeneRNAs, GeneDBs, GeneMIMs, and GeneAccs provide links to external genetic databases;
GeneTisTmp includes information about tissues and EST sources for a gene;
PwNotes, ChemNotes, RONotes, FONotes, and GeneNotes provide links between notes (annotations), pathways, Chems, ReactOrgs, FuncOrgs, and Genes;
Notes includes information about notes and annotations;
PapNote, and Papers provide references for each note;
NoteDiss, and Diseases include information about how diseases are linked to a note, for example, whether a certain entity is thought to be a cause or manifestation of a disease, or is hypothesized to be involved in a disease.
There are multiple ways for elucidation of protein-protein interactions. One approach is to apply text-mining algorithms for screening experimental literature for co-occurrence (therefore, association) of gene/protein symbols and names in the same text. Typically, Natural Language Processing algorithm (NLP) is used for automated mining abstracts and titles of PubMed articles. The reliability of NLP-derived associations can be enhanced by compilation of field-specific synonym dictionaries, using longer word strings for search and full-text articles to query against. In a recent study, the NLP engine MedScan was used to extract 2976 interactions between human proteins from full text articles with a precision of 91% for 361 randomly extracted protein interactions. However, the comparative studies show that, in general, only 30-50% of NLP associations corresponded to experimentally verified protein interactions.
Protein-protein interactions can also be derived from high-throughput experimentation. For example, the yeast 2-hybrid (Y2H) screen test identifies protein pairs capable of dimerization in yeast cells. A widely used wet lab technique, Y2H was scaled-up for global mapping of protein interactions in yeast, fly D. melanogaster and worm C. elegans. Y2H became the technology base for several tools and discovery companies such as Curagen (www.curagen.com) and Hybrigenics (www.hybrigenics.fr). However, Y2H-derived interactions are known for high (over 50%) level of false positives and false negative interactions. The interactions can also be deduced from condition-specific co-occurrence of gene expression based on the assumption that interacting proteins must be expressed in, especially when encoded by the homologous genes. Abundant and readily obtainable even from small cell populations, co-expression-based clustering is thought to become the major source of tissue-, disease- and treatment-specific interactions. However, the overall confidence in co-expression-derived interactions in yeast is about 50% (47% anti-correlation for novel interactions). Another method, co-immunoprecipitation (Co-IP) consists of affine precipitation of protein complexes in mild conditions using antibodies to one of the complex's subunits, followed by mass-spectrometry or Western blot analysis. A true proteomics method, Co-IP was used in back-to-back studies of yeast interactome. The other, less often used experimental and computational methods include protein arrays, fusion proteins, neighbor genes in operons (for prokaryotic proteins), paralogous verification method (PVM), co-localization, synthetic lethality screens and phage display; each method has its merits and biases. The overall confidence in interactions defined as the intersection between interacting pairs obtained with different methods remains dismal. For instance, over 80,000 protein-protein interactions were detected in yeast S. cerevisiae by six high-throughput experimental methods, but only 2,400 of these interactions were supported by more than one method. Such low overlap limits the applicability of direct comparison between HT interactions datasets of different experimental origin. Recently, statistical methods were developed for enhancing the confidence of interactions derived from low confidence data and analyzing the general parameters of interaction datasets. Y2H and Co-IP yeast protein interaction data applied in yeast were extensively compared for experimental biases and correlation. Although only 6% of Y2H interactions were confirmed by Co-IP method, the authors managed to develop a statistical regression model for prediction of biological relevance and confidence of HT interactions based on sub-network analysis. In another study, graph-theoretical statistics were used for comparative analysis of the interaction datasets in yeast. The parameters and algorithms were realized in the publicly available tool TopNet for comparison of biological sub-networks of different origin (networks.gersteinlab.org/genome/interactions/networks/core.html). In general, it is believed that only manually curated physical protein interactions extracted from original small-scale experimental literature can be used with sufficient confidence.
Dozens of the original and compilation academic protein-protein and protein-DNA interaction databases are available, covering high-throughput and small-scale experimental interactions as well as other experimentally and computed interactions. The most relevant and original database projects, pathways database and analytical tools are outlined in Table 5.
Biological networks are presented as nodes (proteins, genes and compounds) connected by edges (protein-protein, protein-gene, protein-compound interactions and metabolic reactions). Depending on the type of underlying data and the interaction mechanism, the edges are either directed or undirected. For instance, protein binding interactions derived from Y2H assays are undirected, while most of the physical interactions extracted from full text articles have one direction (e.g., protein A activates protein B, but not vice versa). There are several major parameters by which networks can be described and compared (
-
- 1) Average degree (K): the average number of edges per node. In directed networks one can distinguish incoming degree (Kin), outgoing degree (Kout) and total degree.
- 2) Average clustering coefficient (43c): the average ratio of the actual number of links between the node's neighbors and the maximum possible number of links between them. Clustering coefficient for the node i can be calculated as Ci=2ni/k(k−1), where ni is the actual number of links connecting k neighbors of the node to each other
FIG. 43 b.
3) Shortest path lAB for the pair of nodes is the minimum number of network edges that need to be passed to travel from A to B. On a directed graph the shortest path from A to B may be different from the path from B to A as shown on
-
- 4) Diameter (D) is the longest distance between a pair of nodes on the graph.
The default random network theory states that pairs of nodes are connected with equal probability and the degrees follow a Poisson distribution. This implies that it is very unlikely for any node to have significantly more edges than average.
The analysis of yeast interactome (the best studied organism in terms of interactions) revealed that the networks are remarkably non-random and the distribution of edges is very heterogeneous, with few highly connected nodes (hubs) and the majority of nodes with very few edges. Such topology is defined as scale-free, meaning that the node connectivity obeys power law: P(k)˜k−γ, where and P(k) is the fraction of nodes in the network with exactly k links. Interestingly, the hubs are predominantly connected to low-degree nodes, a feature that gives biological networks the property of robustness. A removal of even substantial fraction of nodes still leaves the network connected. At the level of global architecture, networks of different origin (e.g. metabolic, regulatory, protein interactions, networks for different organisms) share the same properties. Taken together, the metabolic reactions and signaling interactions form a large cluster linked via molecular nodes shared among many cellular processes. This runs contrary to a traditional model of small and relatively independent linear pathways.
The key property of biological networks is their modular nature. According to modular theory, various types of cellular functionality are provided by relatively small, transient but tightly connected networks of molecules (5-25 nodes) that are engaged in performing specific functions. Identification of such modules is a non-trivial problem as complex networks can be parsed into subsets in many different ways, potentially generating billions of combinations. For example, our analysis of the network of a subset of 35,000 experimentally proven human signaling interactions in the MetaCore™ database revealed about 2 billion linear 5-step network paths, all physically possible. It is clear that only few of these paths are realized in any cell and time as active pathways.
Different approaches have been offered for automated parsing of large networks into modules. One set of methods identifies the modules using various clustering algorithms. These include Monte Carlo optimization methods for finding tightly connected clusters of nodes; clustering based on shortest paths length distribution, and other graph clustering algorithms. It has been shown that some clusters identified in this way do in fact correspond to either known protein complexes or metabolic pathways. Another approach implies analysis of motifs; fairly simple sub-graphs that share certain structural and functional features, such as a feedback or feed-forward loops. The number of different motifs in a given network is calculated and then compared with the number of the same motifs in a randomly connected network. Those motifs in which the network is enriched when compared to the random network may represent potential functional modules. The motifs were identified in regulatory networks of E. coli and yeast. It should be noted that performance of these algorithms is usually judged by how well they can recall the known functional units or processes. On this account, all of these algorithms are prone to a high level of false-positives: the modules not corresponding to any of known pathways.
Conditionally active functional modules can also be elucidated by the analysis of high-throughput molecular data (e.g., gene expression, protein abundance, metabolic profiles) in the context of networks. One straightforward approach relies on statistical clustering of gene expression data followed by mapping the resulting clusters onto the networks obtained from independent sources. The advantage of this approach is the prioritization of gene clusters base on the number of links to the network. The drawback is that the statistics-derived clusters are inherently artificial and can be connected to multiple networks and cellular processes. In another method, the network clustering algorithms such as super-paramagnetic clustering are used to identify tightly connected sets of nodes. The expression data helps to assign weights to the edges and nodes; the combined distance is then computed based on both expression profiles and the network distance between gene products. Other methods include simulated annealing and probabilistic graphical models. Essentially, analysis of molecular data within the context of interaction networks reveals genes/proteins that share a similar pattern of expression and at the same time are closely connected on the network (
The non-random nature of biological networks is associated with biological functions of nodes and edges. Recently, several studies in yeast revealed correlations between the network topology and composition with important biological properties of nodes' proteins. The well-connected hubs (defined here as the top quartile of all nodes in terms of the number of edges) are largely presented by evolutionary conserved proteins as the interactions impose certain structural constrains on sequence evolution. In both yeast S. cerevisiae and worm C. elegans, a significant negative correlation was shown between the number of interactions and the relative evolutionary rate. Recently, it was revealed that the number of interactions positively correlates with essentiality in yeast. Essential and marginally essential (relative importance of a non-essential gene to a cell) genes tend to be hubs with short characteristic path length to the neighbors. Essential proteins tend to be more closely connected to each other. Furthermore, essential proteins tend to be the more promiscuous transcription factors, and the target genes regulated by fewer transcription factors, tend to be essential. Many of these targets are housekeeping genes with high expression levels and less expression fluctuation. It was also noted that soluble proteins feature more interactions than membrane proteins. As mentioned above, the links between highly connected and low-connected pairs of proteins define the specific topology of the networks, characteristic for the condition. In yeast, the direct links between highly connected hubs are suppressed and the hub—low connected node pairs are favored. Such topology probably prevents crosstalk between the functional modules and sub-networks. The findings may have substantial implications for the practice of drug discovery in terms of target prioritization and identification of multi-gene/multi-proteins biomarkers.
Biological networks are the most suitable tool for functional mining of large, inherently noisy experimental datasets such as microarray and SAGE expression patterns, proteomics and metabolomic profiles. There is an important distinction between networks and the other methods available for HT data analysis (such as statistical clustering, linking to pathway databases, process ontology, pathway maps, cross-species comparisons etc.). Unlike other methods, networks' edges provide primary information about physical connectivity between proteins, their subunits, DNA sequences and compounds. The complete set of interactions which assembles into networks on-the-fly, defines the potential of a cell to form multi-step pathways, signaling cascades and protein complexes representing the core machinery of cellular life in health and disease. Obviously, only a fraction of all possible interactions is activated at any given condition as only some of the genes are expressed in tissues at a time and only a fraction of the cellular protein pool is active. The subset of activated (or repressed) genes and proteins are captured by OMICs experiments, such as global gene expression profiles, proteomics or metabolomics profiles—the functional snapshots of cellular response. Analyzed separately, these datasets cannot explain the whole picture. There are many levels of information flow between a gene and an active protein it encodes, including gene expression, mRNA processing, protein trafficking, posttranslational modifications, folding and assembly into active complexes (
Networks represent a flexible and powerful analytical tool for comparison and cross-validation of different types of datasets associated with a condition (disease, drug treatment etc.). In fact, any experimental or literature-derived dataset with recognizable gene or protein IDs (such as LocusLink, Unigene, SwissProt, RefSeq, OMIM) can be visualized, mapped and compared against each other on the same network. For example, one can directly compare the list of genes known from genetics analysis with the gene expression arrays from a patient in clinical trials and a knockout mouse. When the same data type and experimental platform is used, the conditional networks can be compared in great detail for common and different sub-networks and patterns. Such fine mapping can be performed in order to compare the tissue and cell type specific response, different time points, drug dosage; different patients from the same cohort, etc. For instance, we have compared gene expression patterns from mammary gland duct epithelium of two breast cancer patients, one from pre-invasive DSIC stage, another with invasive cancer. Both data sets were used for building the initial networks, and then visualized separately. One of the top-scoring networks included the major cell proliferation activator oncogene c-Myc (
Networks analysis is broadly applicable throughout the drug discovery and development pipeline, both on the biology and the chemistry side. Basically, any type of data which can be linked to a gene, a protein or a compound, can be recognized by input parsers, and subsequently visualized and analyzed on the networks. It makes eligible almost any pre-clinical HT experiment as well as patient DNA or metabolic tests from clinical trials (
-
- Target identification: Experimental data from model organisms, cell lines and human tissues can be uploaded and mapped on networks. New hypotheses can be made on the pathways connecting the proteins of interest.
- Target validation and prioritization: Data cross-referencing on the same networks, maps and pathways.
- Disease biomarkers: The biomarkers can be identified as signature networks—condition-specific conserved sets of nodes supported by differential gene expression and protein abundance data.
- Toxicity biomarkers: Same as above, with signature networks derived from toxicogenomics data—typically a rat or mouse liver arrays from drug-treated animals.
- Pharmacogenomics/haplotyping: The networks modules can be used as a mean for haplotyping SNPs associated with the condition.
- Lead optimization and selection of drug candidates: The biology side of small compounds development deals with prioritization of primary indications, possible side effects and ADME/Tox evaluation of novel compounds. New compounds and their metabolites from pre-clinical studies can be mapped on tissue, disease specific metabolic and regulatory networks via structure similarity search with metabolites and ligands included in the database. This functionality is realized in MetaDrug.
- Clinical studies: The patients data (specific DNA sequences, expression microarrays, metabolites from body fluids) can be mapped on networks and compared with pre-clinical data and published experiments.
- New indications for marketed drugs: Secondary indications is an important part of follow-up development for bioactive compounds. New therapeutic areas can be suggested by analysis of tissue-specific, disease-specific networks from animals and humans treated with the drug.
- Post-market monitoring: The patients' data (usually metabolites from body fluids) can be stored in the database and monitored on the networks built during clinical and pre-clinical studies.
Now, we will consider identification of novel therapeutic targets by reverse engineering the network created around existing drug targets. In this case, we used the software suite MetaCore™ previously developed by GeneGo, Inc. In this example, we have uploaded a list of about 40 proteins known as breast cancer therapeutic targets and used this list to build networks with different algorithms applied at MetaCore (shortest path algorithm is presented here). Most proteins have connected into highly concise networks closely associated with cell proliferation and cell cycle progression. Next, we used these networks for mapping published microarray gene expression data from invasive breast cancer patients (the nodes with red circles). The putative novel targets must satisfy the following conditions: 1) connectivity in one step with the known targets, 2) be upstream of known targets in signaling, and 3) condition-specific overexpression.
The networks can be used in a similar way for identification of biomarkers. In Example 10 (
The following examples are offered to illustrate but not to limit the invention.
Example 1 Stabilization of Heparin for Treatment of ArteriosclerosisHC gp-39, a protein of the chitinase family, can be used in combination with heparin to treat arteriosclerosis. Addition of HC gp-39 may stabilize heparin and increase its effectiveness.
Heparin appears to play a role in arteriosclerosis. Data shows that patients suffering from arteriosclerosis have decreased heparin levels. Therapeutic treatment with heparin is used to reduce the risk of infarction and stroke. Heparin is also used as an anti-coagulant. It activates antithrombin-III. Additionally, low molecular weight heparin is used for the treatment of lipid metabolism disorders as an agent that activates lipoprotein lipase.
Under normal conditions, lipoprotein lipase is localized on the cell surface, including the surface of endothelial cells in blood vessels. The binding of heparan sulfate to lipoprotein lipase is responsible for the retention of lipoprotein lipase on the cell surface. While bound to the cell surface, lipoprotein lipase is not enzymatically active, but serves as a receptor, binding low density and very low density lipoproteins (LDL and VLDL). This binding leads to the cellular uptake of lipoproteins (PMID 10532590). Development of arteriosclerosis is characterized by the emergence of so-called foam cells that form due to an excess of lipoproteins being absorbed into the cell through pinocytosis.
Heparin has a higher affinity for lipoprotein lipase than does heparan sulfate. With the exchange of heparin for heparan sulfate binding to lipoprotein lipase, the lipoprotein lipase is activated and released from the cell surface and into the intercellular space and to the blood (PMID 11427199). While the binding of heparin activates lipoprotein lipase, in the absence of heparin, even if lipoprotein lipase is released from the cell surface, it remains inactive (PMID 10760480).
The binding of heparin to lipoprotein lipase results in several positive therapeutic effects. First, the uptake of lipoproteins by cells is decreased and, therefore, further formation of foam cells is prevented. Second, heparin-bound lipoprotein lipase regains its catalytic activity (PMID 210908, 698674) and starts to degrade LDL and VLDL in the intercellular space and in the blood. An excess of LDL and VLDL in the blood leads to the formation of atherosclerotic plaques. In contrast, degradation of LDL and VLDL by lipoprotein lipase leads to the formation of fatty acids that are eventually processed in the liver. Therefore, the degradation of LDL and VLDL by lipoprotein lipase helps prevent the development of arteriosclerosis.
As mentioned above, patients with arteriosclerosis are often treated with heparin. Free heparin is thought to be degraded by heparinase. A full length human heparinase enzyme has not been isolated. Human heparinase is known only by fragments of its sequences (NCBI protein # AAE10146-10153, ME13758-13770, AAE67749-67785). While the enzymatic activity of human heparinase has not been directly studied, other known heparinases belong to the class of enzymes known as Iyases. Based on similarities to known heparinases, it is likely that human heparinase interacts with heparin through binding to its non-reducing end and degrades heparin.
HC gp-39, a protein of chitinase family, can also bind to heparin (Medline 96325055). The binding of heparin (or heparin analogs) to HC gp-39 may protect heparin from degradation by heparinase (
Currently, there is no direct evidence regarding the way in which HC gp-39 binds to heparin. It is known, however, that some hydrolases, which are close to chitinases, bind their substrates at the non-reducing end of the substrate. HC gp-39, therefore, may similarly bind to the non-reducing end of heparin. This binding would protect heparin from degradation by heparinase. The HC gp-39 homolog from pig smooth muscle culture (porcine gp38k) has been studied in greater detail. HC gp-39 shows 84.6% homology with gp38k (DNAstar). The site of heparin binding on gp38k (residues 144-149, RRDKRH) is similar to a putative heparin binding site on HC gp-39 (RRDKQH) in which glutamine is substituted for arginine in the human protein.
Example 2 Tissue RemodelingIn most tissues, cells are connected through a membrane-based complex of polysaccharides and through membrane-linked proteins known as the glycocalix and the extra-cellular matrix. Heparan sulphate is one of the most important components of both the glycocalix and the extra-cellular matrix. Heparan sulphate binds to fibronectin and other structural proteins; this binding is required for the fixation of cells within tissues and determines tissue structure (
HC gp-39, a protein of the chitinase family, has a higher affinity for heparan sulphate than does fibronectin (Medline 96325055). HC gp-39 may compete with fibronectin for the binding of heparan sulphate. If HC gp-39 binds to heparan sulphate replacing fibronectin, intercellular bonds and the structural components which retain tissue structure can be relaxed. Such relaxing is required for successful tissue remodeling and regeneration. By increasing the local concentration of HC gp-39 and thereby locally relaxing structural elements of a tissue, tissue remodeling and regeneration can be stimulated. Such an application would be useful in such areas as wound healing and joint alterations due to arthritis.
Example 3 ArteriosclerosisHyaluronic acid (HA) binds to smooth muscle cells and prevents their proliferation. Proliferation of smooth muscle cells in arteriosclerosis leads to the growth of the arteriosclerotic plaque. Therefore, HA is a factor that helps contain the disease. Chitotriosidase, or chitinase 1, may restrict the synthesis of HA by degrading the chitin primers necessary for HA formation. Therefore, chitotriosidase facilitates the growth of arteriosclerotic plaques. Suppression of the activity of chitotriosidase may be useful in the treatment of atherosclerosis (see
Hyaluronic acid is involved in various processes of tissue repair and remodeling. In particular, HA plays a role in the regulating the migration and proliferation of smooth muscle cells which are critical in the pathogenesis of cardiovascular diseases. HA acts as a negative regulator of the proliferation of smooth muscle cells induced by platelet-derived growth factor (PDGF) and as a positive regulator of PDGF-induced migration (PMID: 9678773, 8842351, 7568237).
Uncontrolled proliferation of smooth muscle cells facilitates the growth of atherosclerotic plaques. As cells start to actively absorb lipid particles, turning into foam cells, the cells form the core of the plaque. Additionally, proliferation of smooth muscle cells leads to the enlargement of the formation and the isolation of the foam cells by covering them with new layers of smooth muscle cells. This further leads to the formation of atheroma, or the degeneration of the artery lining. Drugs that reduce smooth muscle cell proliferation are often used as a part of atherosclerosis therapy. Most of these drugs, however, are hormones that have many undesirable side effects and may be restricted in their use.
HA is synthesized on the extracellular side of the plasma membrane of various cell types, including smooth muscle cells and endothelial cells (PMID: 10493913). Apparently, fibroblasts provide a source for much of the HA implicated in atherosclerotic damage (see e.g., PMID: 11378333, 11327061, 11171074). HA synthesis is catalyzed by the enzyme hyaluronan synthase (HAS). Presently, three human genes for this enzyme have been identified: HAS-1, HAS 2, HAS 3, mapping to chromosomal regions 1 9q13.3-q13.4, 8q24.12, and 16q22.1, respectively. HAS is a plasma membrane proteins.
It has been shown that human hyaluronan synthase is highly homologous to the enzymes from other organisms including glycosaminoglycan synthase from Xenopus (DG42). (PMID: 8798544, 8798477). It has been shown that DG42 and its analogs from zebrafish and mouse exhibit chitin oligosaccharide synthase activity. Furthermore, addition of purified chitinase to zebrafish cell extracts leads to significant (up to 87%) reduction in the synthesis of HA. Based on these data, it is thought that chitin oligosaccharides serve as primers for hyaluronic acid synthesis (PMID: 8643441).
Chitotriosidase (EC 3.2.1.14) and HC gp-39 expressed by macrophages in the area of atherosclerotic damage have been found in the blood vessel wall matrix. It has been suggested that chitotriosidase recognizes the HA primer as its own substrate and, therefore, interferes with the synthesis of HA (PMID: 10073974).
The mechanism by which chitotriosidase participates in the process of regulating proliferation and migration of smooth muscle cells may be based on its enzymatic activity with respect to chitin-like oligosaccharides that serve as primers for HA synthesis. The cleavage of these primers by chitotriosidase may lower the local concentration of HA, therefore, leading to an increase in cell proliferation causing further damage to the blood vessel wall.
Example 4 CosmeticsGlycosaminoglycans are widely used in dermatology and cosmetology for healing and regeneration of skin damage due to trauma, surgery, or aging. In the past decade, a number of cosmetics and therapeutic treatments; containing glycosaminoglycans were developed and marketed for topical use and for injection. Compositions have included glycosaminoglycans such as chitosan, hyaluronic acid, heparin, heparan sulphate, and others. The inclusion of human lectin HC gp-39 into topical compositions with; glycosaminoglycans may accelerate and prolong skin improvement (
Addition of HA to the extra-cellular matrix causes hydration and increases turgor in a tissue. As discussed above, HA is also one of the; important factors in tissue remodeling, as it interacts with a number of proteins and non-protein components of extra-cellular matrix to form a scaffold for the formation of cell layers. HA stimulates the expression of metal proteases in the extra-cellular matrix, for example, elastase-like endopeptidases expressed in fibroblasts and keratinocytes. Both of these cell types receptors for binding hyaluronic acid which is needed for tissue remodeling.
The use of HC gp-35 in combination with hyaluronic acid, may play a function similar to lectin, having a loosening effect on both protein and; glycosaminoglycan elements of the extra-cellular matrix. Treatment with HC gp-39 and HA would preferably be followed by treatment with fibroblast growth factor (FGF) and insulin-like growth factor (IGF) in order to stimulate expression of HAS1, HAS2 and HAS3 for endogenous synthesis of HA (
Therapeutic or preventive treatment with HA is especially important for elderly patients or patients with age-related conditions because the level of endogenous HA diminishes with age. (With age, the number of lipid-filled macrophages raises causing an increase in the concentration of chitotriosidase and, correspondingly, the depletion of endogenous HA.) HA is also capable of deep penetration into the epidermis and may be used as a vehicle for drug delivery.
Example 5 Parkinson's DiseaseOne of the treatments for Parkinson disease includes transplantation of neurons from the substantia nigra of 6-10 week old embryos. The effectiveness of this treatment depends on the successful incorporation of the transplanted tissue. Currently employed techniques show fairly low success rate. The low success rate is; related to rejection of the transplant, usually within several months after surgery. It has been shown that successful transplantation can be achieved with the addition of embryonic neuro-ectodermal cells of Drosophila melanogaster into the transplant tissue. (PMID: 9532720; PMID: 9449456). These cells are known to express a number of growth factors and remodeling factors, including DS47, which is homologous to human protein HO gp-39.
Incorporation of a transplant is related to the processes of tissue remodeling. Integration of transplanted cells into a damaged tissue and; differentiation of the transplanted cells is necessary for restoring the function of the damaged tissue. These processes are related to tissue remodeling, and remodeling factors play a significant role in the interaction of transplanted cells with the extra-cellular matrix and the cells of the recipient. Often rejection of the transplant is not due to an immune response in the recipient, but rather to the lack of tissue integration caused by the formation of filial scar tissue and the lack of blood vessel in-growth into the transplanted tissue. One apparent reason is the low activity of remodeling factors in the recipient tissue. In particular, the rejection it may be related to age-dependent weakening of remodeling capabilities.
It may be possible to regulate tissue remodeling upon transplantation by changing the local concentration of remodeling factors, including proteins belonging to chitinase family such as HC gp-39. Activity of brain chitinases should be related to microglial cells that are descendants of blood monocytes. Neutral cells of a transplant, on the other hand do not accumulate enough remodeling factors due to their nature. The significant increase in transplant integration success rates by incorporating Drosophila embryonic cells suggests that these cells actively express remodeling factors that are closely related to such factors in humans. It is known that four proteins belonging to the chitinase family are expressed in the human brain (HC gp-39, chitotriosidase, YKL 39, and FLJ12549). There is also expression of chitinase-like proteins in the embryonic cells of Drosophila. These proteins lack catalytic activity, but are capable of binding with proteoglycans of the extra-cellular matrix. One of the Drosophila proteins shows slightly homologous to human HC gp-39 (PMID 7875581).
Example 6 System Reconstruction of Emericella nidulansThis example presents the first study of metabolic reconstruction of a eukaryotic organism based solely on Expressed Sequence Tag (EST) data. As illustrated in the present example, the process of the present invention can be used to study metabolism, not just in humans, but in any species. This study was performed within the framework of the WIT 2 system, a WEB-based environment for comparative analysis of genomes, publicly available at the University of Oklahoma's Advanced Center for Genome Technology. The WIT Project was instituted to develop a framework for the comparative analysis of genomic sequence data, focusing largely on the development of metabolic models for sequenced organisms.
Emericella nidulans (formerly Aspergillus nidulans) was chosen as a model organism for this work. Emericella nidulans has been a classical genetic organism for more than fifty years. Its unique metabolism has been extensively studied, especially with regard to carbon compounds. Carbon and alcohol metabolism, nitrogen assimilation, acetamide and proline utilization, amino acid metabolism, sulfur metabolism, and penicillin and sterigmatocystin biosynthesis are the best characterized metabolic systems in E. nidulans.
Gene expression and regulation have also been studied extensively in E. nidulans. There are some fairly well understood systems, such as nitrogen metabolite repression, carbon catabolite repression, regulation of acetamide utilization, regulation of purine degradation, regulation of metabolic flux in the quinate and shikimate pathways, and regulation of gene expression by pH, oxygen and phosphorus. Recently, significant progress has been made towards understanding genetic regulation of reproduction and development in E. nidulans. See, Adams et al., Coordinate control of secondary metabolite production and asexual sporulation in Aspergillus nidulans. Moreover, Emericella belongs to a family of industrially important fungi, some of whose members are common human opportunistic pathogens, and all of which are able to produce penicillin and carcinogenic toxins (aflatoxin, sterigmatocystin, etc.). The genome size of E. nidulans is about 30 Mb. This organism has a typical ascomycetes life cycle, which includes a vegetative stage and three reproductive cycles: sexual, asexual, and parasexual.
EST data for Emericella nidulans and Neurospora crassa were provided by Oklahoma University. Unigene databases for both organisms were created by multiple sequence alignments of different ESTs which were believed to correspond to the same actual gene, providing a more accurate and longer version of the gene sequence. 4155 unigene ESTs were provided for Emericella nidulans (abbreviated EN in Table 1) and 633 unigene ESTs were provided for Neurospora crassa (abbreviated NC in Table 1).
Using these unigene entries, similarities to known protein sequences were computed using blastx and by comparison to other EST sequences using blastn. The results are summarized in Table 1. The numbers in Table 1 represent the percentage of sequences from E. nidulans and N. crassa that show similarity to sequences from each of the other organisms listed. For example, 29.2% of E. nidulans sequences and 34.9% of N. crassa sequences show similarity to the yeast sequence.
About 40-60% of the sequences fail to show similarity to any protein in the non-redundant protein database with a cutoff of 1.0e, which is quite strict. When the cutoff was set at 1.0e-2, an additional 5% of the ESTs showed recognizable similarity. The fraction of hits against proteins with known function in Emericella nidulans is slightly lower than the percentages that are seen with complete chromosomal sequences for the ORFs, which is about 55-60% at this time). EST data, and even unigene EST data, is made up of relatively short sections of genes that include frameshifts. Without the frameshifts, blastx (or FastA) would produce excellent results. The recognizable similarities would certainly go up in the cases involving frameshifts if they could be corrected or if approximate translations estimating the position of the frameshift could be produced. It may be possible to achieve this type of result if ESTs from a closely related organism were available.
The goal of the instant example is to produce an accurate System Reconstruction for Emericella nidulans based on the available EST data. System Reconstruction generally involves two steps. First, assignment of a function to each unigene number is made. Second, a set of metabolic pathways specific for the organism is identified. Since each asserted pathway is composed of a set of functional roles (i.e., enzymes), the unigene entries, with their appropriate functions and corresponding EC numbers, were associated with each of the asserted pathways. The comparative value of the reconstruction from EST data versus reconstruction based on genomic data is summarized in Table 2 below.
Assignments were made to about 2,800 of the ESTs, and then development of an emerging model of the metabolism of E. nidulans began. An extensive literature search for E. nidulans has been performed. The search focused on known metabolic pathways of this organism, as well as on gene regulation and physiology of filamentous fungi. Almost every pathway asserted for E. nidulans has a corresponding reference included in the annotation. The current reconstruction is composed of more than 600 asserted pathways which connect to about 500 specific ESTs. Many pathways are composed of a single reaction, and many others are known to exist biochemically but specific ESTs corresponding to the appropriate functional roles could not be identified. Thus, the collection of assigned functions and asserted pathways represents a model of the metabolism of E. nidulans. This model can be integrated with the growing body of both genetic sequence data and available biochemical characterizations. Such integration forms the basis for a continuing analysis of the organism. The current status of system reconstruction for both S. cerevisiae and E. nidulans is summarized in Table 3 below. Some of the asserted pathways have broken down into categories. The numbers in Table 3 indicate where the analysis is relatively complete and where it is sparse or lacking altogether. Some of these pathways are single reactions that may have similar forms in different cell states.
As the System Reconstruction of E. nidulans for a given number of unigene entries was completed, a visual outline for major parts of metabolism was created. Such schemes not only provide descriptive overviews of certain parts of metabolism, but also reflect the expression patterns specific for a given EST library. The expression patterns become evident when the representation of enzymes in pathways is compared with different sources of expression data, independent from EST data. The expression pattern of identified genes in the reconstruction strongly correlates with data present in the literature, further validating the method of System Reconstruction. For example, one of the most important secondary metabolic pathways, the sterigmatocystin biosynthetic pathway, composed of at least 29 enzymatic activities, is developmentally regulated. A positive correlation between both asexual and sexual sporulation and synthesis of the mycotoxin has been documented. In the present study, a cDNA library was constructed from E. nidulans, strain FGSC A26 (veA 1, bio), which had undergone development for 24 hours on a solid surface with an air interface and, therefore, contained cDNAs from both vegetative mycelial cells and cells involved in asexual reproduction. Indeed, unigene numbers for all 29 genes in the pathway have been identified, and most of them had several candidates for the same gene. Another example is the penicillin biosynthetic pathway which consists of only 3 enzymes: DELTA-(L-ALPHA-AMINOADIPYL)-LCYSTEINYL-D-VALINE SYNTHETASE (acvA), ISOPENICILLIN N SYNTHETASE (ipnA), and ACYL-COENZYME A:6-AMINOPENICILLANIC ACID ACYLTRANSFERASE (aatA). Expression of both acvA and aatA is slightly repressed by glucose in fermentation medium. Consistent with literature data, there are no unigene candidates for acvA, one for aatA, and two for ipnA.
The reconstruction of E. nidulans metabolism illustrates the use of System Reconstruction from EST data. In fact, alterations to WIT required to support an analysis based upon both EST and chromosomal sequence data have been made. The outcome represents an initial effort to encode the known metabolism of E. nidulans and to relate the analysis to actual sequence data (in this case largely ESTs). Such an effort lays the foundation for an ongoing analysis of the genome and embeds the analysis in a framework that supports comparative analysis between organisms.
Example 7 System Reconstruction of Amino Acid MetabolismThe System Reconstruction method was used to analyze amino acid metabolism in humans. A portion of the reconstructed map showing the TCA cycle is shown in
For each of the enzymes, the following data is collected: systematic name and synonyms; EC number (if assigned); a spectrum of substrates and products, including not only specific compounds, but also classes of compounds; known inhibitors and activators; kinetic data, including constants such as KM and Vmax for the enzyme or semi-quantitative data on reaction time-scales; and bibliographic references.
The database of amino acid metabolism includes about 150 reactions and pathways described in biomedical literature as involved in biosynthesis and degradation of amino acids. These are reactions and pathways that have been identified experimentally. The following types of information are collected for each reaction or pathway: participating compounds and their roles; a spectrum of enzymes catalyzing the reactions in the pathway, indicating enzymes whose involvement has been identified experimentally in vivo and, those that could participate in the pathways based on their ability to catalyze pathway's reactions; localization and compartmentalization of components; kinetic data, whenever available; and bibliographic references.
For intermediate compounds that occur in the collected pathways and reactions, the following types of data are collected: systematic name of the compound and synonyms; compound classification and compound major structural and functional groups; the endogenous status of the compound in human metabolism (whether the compound occurs as a natural intermediate in human metabolism); thermodynamic data such as free energy, enthalpy and entropy of formation; and bibliographic references. Thermodynamic data are used in combination with metabolic profiles to evaluate the plausibility of the proposed novel pathways.
The first step in building functional models is to link the collected pathways into metabolic networks. There are different types of molecules as well as different types of interactions between biological molecules, and these are indicated through different types of links. Such links are implicitly contained in the database. Indeed, whenever two pathway records share a common intermediate, or an intermediate in one pathway occurs as a regulatory factor in a record for the enzyme from another pathway, it implies a link between these two pathways. Further computations would be facilitated, however, if such links translate into explicit relations among pathways. To this end, a set of special database queries have been developed that extract such relationships and generate tables to describe such links explicitly. These tables constitute a computer representation of a biochemical network that forms a skeleton of the System Reconstruction Model. Unlike the assembled or statistically inferred networks used in many studies, the System Reconstruction Model is built from experimentally verified pathways that may be thought of as identified routes on a biochemical network. It is important to note that only a small fraction of all possible reaction sequences are realizable as functional pathways in any given organism. The types of relationships included in the network may include, for example, the following: pathways linked by shared substrates and/or products; activation of an enzyme by the intermediate metabolite; inhibition of an enzyme by the intermediate metabolite; metabolites that lead to the induction of expression of an enzyme-related gene; metabolites that lead to the suppression of the expression of a gene; and regulation of a transporter or channel by an intermediary metabolite. As the data are collected, other import links may become evident and can be included in the model.
The next step involves converting the network of pathways into a System Model. A network of pathways is only a skeleton on which other data can be assembled. Data integration is accomplished by a specially developed procedure called Structured Annotation. In the course of this procedure, links are established between particular elements in a pathway network. Elements include, for example, pathways, enzymes, metabolites, and the like. This procedure is practically achieved by filling in the annotation tables associated with each element. There are three major categories of data that are integrated into the model at this stage: function-related information; molecular data; and clinical manifestations of human diseases.
Function-related information for pathways and reactions includes functional roles in the human body. These roles may be represented as the catabolism or biosynthesis of certain important molecules, cell energetics, activation, inhibition of various cellular processes, and the like. Functional assignments are not exhaustive, as they have likely resulted from the sets of experiments focused on the specific function. Taken together and integrated within the network of pathways, however, they represent a useful picture of biological functionality and its underlying mechanisms.
The types of information used include organ and tissue localization of the pathway element; intracellular localization and/or compartmentalization; the existence and subcellular localization of the element in other organisms; and references to the primary information source.
Molecular data may include, for example, sequence data, such as genes, ORFs, and Unigene clusters that are associated with enzymes; conditional expression information for an enzyme; genetic polymorphisms of an enzyme and the impact of such polymorphisms on its properties; references to the primary information source; cross-references to records in public genomic databases such as Genebank and TrEMB1; and the like.
Clinical manifestations may include, for example, connection of the element with a disorder (cause, manifestation, and the like), references to the primary information source, and the like. One feature of the model is the incorporation of clinical manifestations (traits) and the ability to view and analyze these data types within the framework of other data integrated into the model. Some clinical traits are directly linked to alteration of a certain biological functions while others are associated with particular genes, proteins, or compounds. The latter are often statistical correlations (e.g., a mutation in a gene correlates with predisposition to a certain disease). In the System Reconstruction Model, biological functions, molecular data, and clinical traits are all linked to a network of pathways. Such a representation allows for the elucidation of the biochemical mechanisms that underlie specific clinical observations.
The user interface of the reconstruction is an interactive map (
The serine biosynthesis scheme, illustrated in
Details for each reaction in the pathway also are accessible from the scheme page. In the serine biosynthesis scheme (
From the scheme page (
Links to nucleic acid sequences and related literature are also available from the enzyme pages. For example, from the enzyme page for EC 3.1.3.3, phosphoserine phosphatase, shown in
Examples of sequences linked to the enzyme page (
The System Reconstruction method used to analyze amino acid metabolism in humans, as discussed in Example 7, allowed the elucidation of a number of previously unidentified metabolic links. One such example is related to Parkinson's disease. As illustrated in
By clicking on the link for Parkinson's disease from the phenylalanine catabolism portion of the interactive metabolic map (
The metabolic map for Parkinson's disease shows the mechanism by which L-DOPA metabolism is linked to a respiratory pathway (via 1.6.5.3). Deficiencies in L-DOPA metabolism have long been known as one of the causes of Parkinson's disease. The involvement of the respiratory pathway is, however, a recent discovery. This illustrates one example of how of linkages are determined through the method of System Reconstruction.
As illustrated by the foregoing examples, System Reconstruction provides a highly interactive visual overview of metabolism as well as easy access to an abundant amount of information related to the metabolic pathways in question.
Example 9 Network AnalysisThe method consists of network analysis of multiple experimental datasets relevant to the diseases. We applied the commercial systems biology platform MetaCore (GeneGo, Inc., St. Joseph, Mich.) as a source of protein-protein interactions and as the means for building and visualization of the networks. The workflow proceeds as following:
-
- A small experiment dataset of genes relevant for the disease is compiled from the literature data. The genes in the dataset have been shown associated with the disease pathology in genetic and biochemical experiments. The exact mechanisms for genes involvement in the disease may or may not be known. Small experiments datasets can include SNP and mutation data, gene amplification data due to chromosomal rearrangements; genes identified by family analysis and other genetic analysis methods. We consider such list as a seed dataset for building the initial networks
- The second, high-throughput dataset is defined as a list of genes or proteins with changed expression or abundance in the disease condition. These genes and proteins are typically identified in high-throughput experiments such as global gene expression profiling with DNA microarrays, proteomics or metabolomics methods. Such list can also include the genome-wide SNP maps.
- The third, analytical dataset is defined as the table of protein-protein, protein-DNA and protein-compound interactions characteristic for human. The interactions are extracted from the experimental literature in scientific journals. Currently, MetaCore contains 45,000 of such interactions, including metabolic reactions and signaling interactions. Only experimentally proven interactions are included in the dataset.
- The seed gene lists are uploaded into MetaCore followed by building the initial networks by the standard algorithms in the software (
FIGS. 47 , 48). - The high-throughput dataset is mapped (superpositioned) on the initial networks by the standard Select Experiments tool in MetaCore as shown on
FIG. 47 . - The differentially expressed genes (or differentially abundant proteins) are identified on the same networks as the corresponding nodes connected directly or in two steps to the nodes from the seed dataset.
- Therefore, the whole list of all high-throughput data is narrowed down to several genes (proteins), most relevant to the seed dataset. The differentially expressed genes connected in one step to the genes from seed list are considered as the most likely candidates for drug targets and molecular biomarkers.
- The final list of genes includes the genes from genetics list and the list of over-expressed genes (highly abundant proteins). Direct interactions algorithm is used for building the final network. This network represents the most relevant network for the disease/condition based on the initial three datasets. The over-expressed genes (abundant proteins) on this network are considered as the most likely targets and biomarkers for the disease (condition).
- Further network analysis to determine specific implication of the selected genes.
As a source of an independent, non-expression dataset, we have compiled a list of 51 genes shown to be associated with glaucoma pathology from small-scale, mostly genetics, experiments (Table 4). We named this dataset as the genetics list. The Direct Interactions algorithm allowed to connect 13 of these genes into a concise network (
Only six genes of 51 small scale dataset were common with the set of 496 of differentially expressed genes from microarrays: the over-expressed MMP-1, APOE and c-Fos and under-expressed ENPP1, MMPI, and SLC4A4. Such small direct overlap, typical for is not sufficient for any functional interpretation, rather that the gene lists are inconsistent.
On the next step, we identified the differentially expressed genes in the closest interactions proximity to the core of small experiments set as the most relevant set of differentially expressed genes to the small experiments set. The Analyzed Networks algorithm was applied to the small experiments set and the network built. The cluster size was limited to 50 objects, and only highest confidence interactions mechanisms allowed. The resulted networks were sorted based on z-scores (see above) and 20 top networks with z-scores from 38 to 56 chosen for the analysis.
The z-scores reflect the relative saturation of the networks with the root objects; in this case with the genes from the genetics list. On each out of top 20 networks, at least 40% of the objects were root objects from the genetics list. The networks included two to six differentially expressed genes, connected with the small experiments genes in one or two steps (
In the next step, we combined the genetics list with the list of 32 differentially expressed genes identified at the previous steps. Six genes were common between the lists. The resulted list of 78 genes was used as root objects for building the final Direct Interactions network. A surprising high number of objects, 46 formed one concise network which included 24 nodes from the genetics list, 18 overexpressed genes and 5 down-regulated nodes (
We evaluated the specificity and non-randomness of the final network. First, sets of 78 objects randomly selected from the relevant dataset (the known gene content of Affymetrix microarray recognized at MetaCore networks) were run 500 times as described above. The p-value of the resulted network was 0.99. Second, we added the list of 32 of most highly expressed genes from the dataset to the genetics set and built networks with the same Direct Interactions algorithm. The resultant network contained 15 nodes total, which is non-essentially more than the genetics network itself.
Example 11 Potential Drug Targets. Drugs and Preventive Therapy for Glaucoma Based on Network AnalysisSmall molecules, siRNA and antibody inhibitors of Caspases 1, 4 and 8 may be utilized as therapy for glaucoma. (see
Small molecules, siRNA and protein modulators of human vitamin D receptor may be identified as therapy for glaucoma. Networks show that VDR—vitamin D receptor is connected to and initiates all major hubs on glaucoma-related networks. VDR is over-expressed in glaucoma.
Small molecules, siRNA and antibody inhibitors of MAPK10 kinase may be identified as therapy for glaucoma. MAPK10—map kinase 10—activates AP-1 (c-Jun/c-fos) transcription factor, NF-kb. It is over-expressed in glaucoma.
Small molecules, siRNA and protein inhibitors of the proteins involved in inflammatory response in glaucoma may be identified such as GRO-alpha, CD40L, Clusterin, CD14, IL-8, Toll-like receptors (TLR1). All of these genes implicated in pro-inflammatory and anti-inflammatory responses are all over-expressed in glaucoma (see
Small molecules, siRNA and protein modulators for the proteins implicated in membrane homeostasis and cell adhesion APOD, HDL, SVIL, Actinin—all of which are over-expressed in glaucoma and may be identified using the present system (see
Genes, involved in hereditary neurodegenerative disorders CLN3, CLN2, CLN5 and Galactosylceramidase are all slightly over-expressed in glaucoma (1.5-1.9 times) and small molecules, siRNA or protein inhibitors or modifiers may be identified using the present system (see
Defects in CLN3 are a cause of Batten Disease (BD) (also known as juvenile-onset neuronal ceroid lipofuscinosis type 3; JNCL), a recessively inherited neurodegenerative disorder of childhood, characterized by progressive loss of vision, seizures and psychomotor disturbances. Biochemically the disease is characterized by lysosomal accumulation of hydrophobic material, mainly ATP synthase subunit C. Clinical onset is usually from five to ten years of age. No treatment is available and BD is usually fatal within a decade. The incidence is estimated at 1/20000 to 1/100000 live birth, making it one of the most common neurodegenerative diseases of childhood.
Defects in CLN5 are the cause of Finish variant late-infantile neuronal ceroid lipofuscinosis (VLINCL, also known as ceroid lipofuscinosis neuronal 5 (CLN5), a fatal childhood neurodegenerative disease characterized by progressive visual and mental decline, motor disturbance, epilepsy and behavioral changes. The first symptom is motor clumsiness, followed by progressive visual failure, mental and motor deterioration and later by myoclonia and seizures.
Defects in CLN2 are the cause of classical late-infantile neuronal ceroid lipofuscinosis (LINCL), also known as ceroid lipofuscinosis neuronal 2 (CLN2), a fatal childhood neurodegenerative disease characterized by progressive visual and mental decline, motor disturbance, epilepsy and behavioral changes. The three main subtypes of childhood NCLS defined by the age of onset, clinical features and ultrastructural morphology are infantile NCL (INCL), classical late-infantile NCL (LINCL), or juvenile NCL (JNCL), although a number of other distinct variant forms have been described. Catalytic activity occurs with the release of an N-terminal tripeptide from a polypeptide. Detected in all tissues examined with highest levels in heart and placenta and relatively similar levels in other tissues.
Defects in GALC in the brain are the cause of globoid cell leukodystrophy (GLD, or Krabbe disease). This autosomal recessive disorder deficiency results in the insufficient catabolism of several galactolipids that are important in the production of myelin. Clinically the most frequent form is the infantile form. Most patients (90%) present before six months of age with irritability, spasticity, arrest of motor and mental development, and bouts of temperature elevation without infection. This is followed by myoclonic jerks of the arms and legs, oposthotonus, hypertonic fits and mental regression which progresses to a severe decerebrate condition with no voluntary movements and death from respiratory infections or cerebral hyperpyrexia before two years of age. However, a significant number of cases with later onset, presenting with unexplained blindness, weakness, and/or progressive motor and sensory neuropathy that can progress to severe mental incapacity and death, have been identified.
Defects in GALC in skin fibroblasts, which belongs to family 59 of glycosyl hydrolases show the highest level of activity in testes compare to brain, kidney, placenta, and liver and can also be found in urine.
In the testes galactosylceramidase hydrolyzes the galactose ester bonds of galactosylceramide, galactosylsphingosine, lactosylceramide, and monogalactosyldiglyceride. It is an enzyme with very low activity responsible for the lysosomal catabolism of galactosylceramide, a major lipid in myelin, kidney and epithelial cells of the small intestine and colon. It has an optimal pH between 4.0 and 4.4. Activity is lost when heated at 52 degrees Celsius for five minutes.
In the placenta two forms of galactosylceramidase are produced by alternative splicing.
Example 12 Hormone Therapy in GlaucomaIndividual or combined application of parathyroid hormone, androgen, estrogen and progesterone may be utilized to treat glaucoma. PTHrP may play a protective role in glaucoma.
Androgen, estrogen and progesterone should play protective role in glaucoma. ANDR, ESTR and progesterone receptors are significantly up-regulated protective role in glaucoma.
Having thus described several aspects of at least one embodiment of this invention, it is to be appreciated that various alterations, modifications and improvements will readily occur to those skilled in the art. Such alterations, modifications, and improvements are intended to be part of this disclosure, and are intended to be within the spirit and scope of the invention.
Claims
1. A method for identifying drug or gene therapy targets comprising:
- a) collecting data regarding an organism's metabolism for both non-disease and disease states;
- b) linking the data into metabolic pathways;
- c) identifying the interconnections between the metabolic pathways;
- d) creating a map of the organism's metabolism by integrating information obtained in a), b) and c); and
- e) identifying drug targets by comparing differences between non-disease and disease states using the map.
Type: Application
Filed: Aug 6, 2015
Publication Date: Nov 26, 2015
Applicant: GENEGO, INC. (Saint Joseph, MI)
Inventors: Tatiana NIKOLSKAYA (Portage, IN), Andrej BUGRIM (St. Joseph, MI), Yuri NIKOLSKY (Del Mar, CA)
Application Number: 14/820,469
