Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for predicting biological functions of proteins. In one aspect, a method comprises: obtaining data defining a sequence of amino acids in a protein; processing the data defining the sequence of amino acids in the protein using a neural network, wherein: the neural network is a convolutional neural network comprising one or more dilated convolutional layers; and the neural network is configured to process the data defining the sequence of amino acids in the protein in accordance with trained parameter values of the neural network to generate a neural network output characterizing at least one predicted biological function of the sequence of amino acids in the protein; and identifying the predicted biological function of the sequence of amino acids in the protein using the neural network output.

Abstract: A method for detecting false positive variant calls in a next generation sequencing analysis pipeline involves obtaining a plurality of read pileup windows associated with a first sample genome. The method also involves obtaining, for each reference nucleotide position represented in the plurality of read pileup windows, a label indicating that the reference nucleotide position is either (i) a known variant or (ii) a non-variant. The method further involves training a neural network based on data indicative of the plurality of read pileup windows and the labels. Additionally, the method involves receiving a read pileup window associated with a second sample genome. Further, the method involves determining, using the trained neural network, a likelihood that the read pileup window associated with the second sample genome represents a variant.

Abstract: A method for detecting false positive variant calls in a next generation sequencing analysis pipeline involves obtaining a plurality of read pileup windows associated with a first sample genome. The method also involves obtaining, for each reference nucleotide position represented in the plurality of read pileup windows, a label indicating that the reference nucleotide position is either (i) a known variant or (ii) a non-variant. The method further involves training a neural network based on data indicative of the plurality of read pileup windows and the labels. Additionally, the method involves receiving a read pileup window associated with a second sample genome. Further, the method involves determining, using the trained neural network, a likelihood that the read pileup window associated with the second sample genome represents a variant.

Abstract: The present disclosure is directed to systems and method to perform improved detection of out-of-distribution (OOD) inputs. In particular, current deep generative model-based approaches for OOD detection are significantly negatively affected by and struggle to distinguish population level background statistics from semantic content relevant to the in-distribution examples. In fact, such approaches have even been experimentally observed to assign higher likelihood to OOD inputs, which is opposite to the desired behavior. To resolve this problem, the present disclosure proposes a likelihood ratio method for deep generative models which effectively corrects for these confounding background statistics.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for predicting biological functions of proteins. In one aspect, a method comprises: obtaining data defining a sequence of amino acids in a protein; processing the data defining the sequence of amino acids in the protein using a neural network, wherein: the neural network is a convolutional neural network comprising one or more dilated convolutional layers; and the neural network is configured to process the data defining the sequence of amino acids in the protein in accordance with trained parameter values of the neural network to generate a neural network output characterizing at least one predicted biological function of the sequence of amino acids in the protein; and identifying the predicted biological function of the sequence of amino acids in the protein using the neural network output.

Abstract: A method for detecting false positive variant calls in a next generation sequencing analysis pipeline involves obtaining a plurality of read pileup windows associated with a first sample genome. The method also involves obtaining, for each reference nucleotide position represented in the plurality of read pileup windows, a label indicating that the reference nucleotide position is either (i) a known variant or (ii) a non-variant. The method further involves training a neural network based on data indicative of the plurality of read pileup windows and the labels. Additionally, the method involves receiving a read pileup window associated with a second sample genome. Further, the method involves determining, using the trained neural network, a likelihood that the read pileup window associated with the second sample genome represents a variant.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for processing a biological sequence using a neural network. One of the methods includes obtaining data identifying a biological sequence; generating, from the obtained data, an encoding of the biological sequence; processing the encoding using a deep neural network, wherein the deep neural network is configured through training to process the encoding to generate a score distribution over a set of biological labels for the biological sequence; and classifying the biological sequence using the score distribution.

Abstract: A method for variant calling in a next generation sequencing analysis pipeline involves obtaining a plurality of sequence reads that each include a nucleotide aligned at a nucleotide position within a sample genome. The method also involves obtaining a plurality of alleles associated with the nucleotide position. The method further involves determining that a particular allele of the plurality of alleles matches one or more sequence reads of the plurality of sequence reads, wherein the particular allele is located at the nucleotide position. Additionally, the method involves generating an image based on information associated with the plurality of sequence reads. Further, the method involves determining, by providing the generated image to a trained neural network, a likelihood that the sample genome contains the particular allele. The method may also involves providing an output signal indicative of the determined likelihood.