Abstract: An example method for training a graph convolutional neural network (GCNN) configured for virtual screening of molecules for drug discovery is described herein. The method can include receiving a first data set including a plurality of molecules, and training the GCNN to initialize one or more parameters of the GCNN using the first data set. The method can also include receiving a second data set including a plurality of molecules and respective inhibition rates for a disease, and training the GCNN to refine the one or more parameters of the GCNN using the second data set. The molecules in the first and second data sets can be expressed in a computer-readable format. An example method for virtually screening molecules on Plasmodium falciparum (P. falciparum) is also described herein.

Abstract: Methods, systems, compositions, and computer program products are provided for accurately identifying candidate neoantigens that exhibit imnmunogenic properties. In some embodiments, a method provided herein includes receiving a set of candidate peptide sequences, each candidate peptide sequence in the set having a major histocompatibility complex (MHC) presentation score meeting a pre-set criterion. The method further includes identifying a corresponding MHC peptide sequence associated with each candidate peptide sequence in the set; generating immunogenicity input vectors from the set of candidate peptide sequences by processing a representation of each candidate peptide sequence in the set of candidate peptide sequences and a representation of the corresponding MHC peptide sequence for each candidate peptide sequence in the set.

Abstract: An example method for training a graph convolutional neural network (GCNN) configured for virtual screening of molecules for drug discovery is described herein. The method can include receiving a first data set including a plurality of molecules, and training the GCNN to initialize one or more parameters of the GCNN using the first data set. The method can also include receiving a second data set including a plurality of molecules and respective inhibition rates for a disease, and training the GCNN to refine the one or more parameters of the GCNN using the second data set. The molecules in the first and second data sets can be expressed in a computer-readable format. An example method for virtually screening molecules on Plasmodium falciparum (P. falciparum) is also described herein.