Systems and methods for classifying, prioritizing and interpreting genetic variants and therapies using a deep neural network

Abstract

Described herein are systems and methods that receive as input a DNA or RNA sequence, extract features, and apply layers of processing units to compute one ore more condition-specific cell variables, corresponding to cellular quantities measured under different conditions. The system may be applied to a sequence containing a genetic variant, and also to a corresponding reference sequence to determine how much the condition-specific cell variables change because of the variant. The change in the condition-specific cell variables are used to compute a score for how deleterious a variant is, to classify a variant's level of deleteriousness, to prioritize variants for subsequent processing, and to compare a test variant to variants of known deleteriousness. By modifying the variant or the extracted features so as to incorporate the effects of DNA editing, oligonucleotide therapy, DNA- or RNA-binding protein therapy or other therapies, the system may be used to determine if the deleterious effects of the original variant can be reduced.

What is claimed is: 
     
         1 . A computer-implemented method for computing a set of variant-induced changes in a condition-specific cell variable for a genetic variant, comprising processing a set of variant features using a cell variable predictor to quantify a condition-specific variant cell variable without obtaining a reference measurement of the genetic variant across a plurality of conditions. 
     
     
         2 . The method of  claim 1 , wherein the genetic variant comprises a variant in a deoxyribonucleic acid (DNA) or ribonucleic acid (RNA) variant sequence relative to a DNA or RNA reference sequence, and wherein the method further comprises extracting the set of variant features from the DNA or RNA variant sequence. 
     
     
         3 . The method of  claim 1 , further comprising extracting a set of reference features from the DNA or RNA reference sequence, and processing the set of reference features using the cell variable predictor to quantify a condition-specific reference cell variable. 
     
     
         4 . The method of  claim 3 , wherein the set of variant features is extracted from the DNA or RNA variant sequence by generating:
 a. a binary matrix with 4 rows and a number of columns equal to a length of the DNA or RNA variant sequence or the DNA or RNA reference sequence, wherein each column contains a bit indicating the nucleotide value at the corresponding position in the DNA or RNA variant sequence or the DNA or RNA reference sequence;   b. a set of features computed using one or more layers of an autoencoder other than the input and output layers of the cell variable predictor; or   c. a set of features that correspond to one or more of: RNA secondary structures, nucleosome positions, and retroviral repeat elements.   
     
     
         5 . The method of  claim 3 , further comprising computing, using the cell variable predictor, probabilities for discrete levels of the condition-specific cell variable, wherein each of the set of variant-induced changes is computed by:
 a. summing an absolute difference between the computed probabilities for the condition-specific reference cell variable and the condition-specific variant cell variable;   b. summing a Kullback-Leibler divergence between the computed probabilities of the condition-specific reference cell variable and the condition-specific variant cell variable for each condition; or   c. computing an expected value of the condition-specific reference cell variable and the condition-specific variant cell variable, and subtracting the expected value of the condition-specific reference cell variable from the expected value of the condition-specific variant cell variable.   
     
     
         6 . The method of  claim 1 , wherein the cell variable predictor comprises a deep neural network. 
     
     
         7 . The method of  claim 6 , wherein the deep neural network comprises a convolutional neural network, a recurrent neural network, or a long-term short-term memory recurrent neural network. 
     
     
         8 . The method of  claim 1 , further comprising combining the set of variant-induced changes in the condition-specific cell variable to compute a single numerical variant score for the genetic variant, the single numerical variant score computed by:
 a. outputting the score for a fixed condition;   b. summing the variant-induced changes across a plurality of conditions; or   c. computing the maximum of the absolute variant-induced changes across a plurality of conditions.   
     
     
         9 . The method of  claim 1 , further comprising computing, for a pair of genetic variants, a distance between the two genetic variants in the pair by summing the output of a nonlinear function applied to a difference between the change in the condition-specific cell variable for the first of the two genetic variants and the change in the condition-specific cell variable for the second of the two genetic variants. 
     
     
         10 . The method of  claim 1 , wherein the genetic variant comprises a) two or more distinct single nucleotide variants (SNVs); or b) a combination of substitutions, insertions, and deletions, wherein the combination is not a single nucleotide variant (SNV). 
     
     
         11 . A computer-implemented method for computing a set of variant-induced changes in a condition-specific cell variable for a genetic variant, comprising processing a set of variant features using a cell variable predictor to quantify a condition-specific variant cell variable, wherein the cell variable predictor comprises a deep neural network comprising at least two connected layers of processing units. 
     
     
         12 . The method of  claim 11 , wherein the genetic variant comprises a variant in a deoxyribonucleic acid (DNA) or ribonucleic acid (RNA) variant sequence relative to a DNA or RNA reference sequence, and wherein the method further comprises extracting the set of variant features from the DNA or RNA variant sequence. 
     
     
         13 . The method of  claim 11 , further comprising extracting a set of reference features from the DNA or RNA reference sequence, and processing the set of reference features using the cell variable predictor to quantify a condition-specific reference cell variable. 
     
     
         14 . The method of  claim 13 , wherein the set of variant features is extracted from the DNA or RNA variant sequence by generating:
 a. a binary matrix with 4 rows and a number of columns equal to a length of the DNA or RNA variant sequence or the DNA or RNA reference sequence, wherein each column contains a bit indicating the nucleotide value at the corresponding position in the DNA or RNA variant sequence or the DNA or RNA reference sequence;   b. a set of features computed using one or more layers of an autoencoder other than the input and output layers of the cell variable predictor; or   c. a set of features that correspond to one or more of: RNA secondary structures, nucleosome positions, and retroviral repeat elements.   
     
     
         15 . The method of  claim 13 , further comprising computing, using the cell variable predictor, probabilities for discrete levels of the condition-specific cell variable, wherein each of the set of variant-induced changes is computed by:
 a. summing an absolute difference between the computed probabilities for the condition-specific reference cell variable and the condition-specific variant cell variable;   b. summing a Kullback-Leibler divergence between the computed probabilities of the condition-specific reference cell variable and the condition-specific variant cell variable for each condition; or   c. computing an expected value of the condition-specific reference cell variable and the condition-specific variant cell variable, and subtracting the expected value of the condition-specific reference cell variable from the expected value of the condition-specific variant cell variable.   
     
     
         16 . The method of  claim 11 , wherein the deep neural network comprises a convolutional neural network, a recurrent neural network, or a long-term short-term memory recurrent neural network. 
     
     
         17 . The method of  claim 11 , further comprising combining the set of variant-induced changes in the condition-specific cell variable to compute a single numerical variant score for the genetic variant, the single numerical variant score computed by:
 a. outputting the score for a fixed condition;   b. summing the variant-induced changes across a plurality of conditions; or   c. computing the maximum of the absolute variant-induced changes across a plurality of conditions.   
     
     
         18 . The method of  claim 11 , further comprising applying thresholds that are fixed or selected using labeled data to the single numerical variant score for the genetic variant to classify the genetic variant (i) as one of deleterious or non-deleterious, (ii) as one of pathogenic, likely pathogenic, unknown significance, likely benign, or benign, or (iii) using another discrete set of labels. 
     
     
         19 . The method of  claim 11 , further comprising computing, for a pair of genetic variants, a distance between the two genetic variants in the pair by summing the output of a nonlinear function applied to a difference between the change in the condition-specific cell variable for the first of the two genetic variants and the change in the condition-specific cell variable for the second of the two genetic variants. 
     
     
         20 . The method of  claim 11 , wherein the genetic variant comprises a) two or more distinct single nucleotide variants (SNVs); or b) a combination of substitutions, insertions, and deletions, wherein the combination is not a single nucleotide variant (SNV).