Methods, systems, and computer readable media for evaluating variant likelihood

Abstract

A method comprises receiving an ensemble of sequencing reads based on measurements from a plurality of microwells of a sensor array; assigning measured values to the ensemble of sequencing reads; calculating model-predicted values utilizing a predictive model of nucleotide incorporations resulting from flows of nucleotide species according to a predetermined order; modifying at least some model-predicted values using a first bias for forward strands and a second bias for reverse strands, the modifying based on variations between model-predicted values for different hypothesized sequences obtained using the predictive model of nucleotide incorporations resulting from the flows of nucleotide species according to the predetermined order; calculating a measurement confidence value for each read in the ensemble of sequencing reads, the confidence value representing variations between the measured values and the modified model-predicted values; and identifying a plurality of reads in the ensemble as corresponding to a variant sequence.

What is claimed is: 
     
         1 . A method of identifying sequence variants, the method comprising:
 receiving an ensemble of sequencing reads based on measurements obtained from a plurality of microwells of a sensor array, the measurements indicative of incorporation of nucleotide species into a plurality of template polynucleotide strands disposed in the plurality of microwells of the sensor array;   assigning measured values to the ensemble of sequencing reads;   calculating model-predicted values for the plurality of template polynucleotide strands utilizing a predictive model of nucleotide incorporations resulting from flows of nucleotide species according to a predetermined order;   modifying at least some model-predicted values using a first bias for forward strands and a second bias for reverse strands, wherein the modifying is based on variations between model-predicted values for different hypothesized sequences obtained using the predictive model of nucleotide incorporations resulting from the flows of nucleotide species according to the predetermined order;   calculating a measurement confidence value for each read in the ensemble of sequencing reads, wherein the measurement confidence value represents variations between the measured values and the modified model-predicted values; and   identifying a plurality of reads in the ensemble as corresponding to a variant sequence based on the measurement confidence value.   
     
     
         2 . The method of  claim 1 , wherein the modifying comprises applying a transformation including a product of one of the first bias and the second bias and a discriminant vector representing a difference between vectors of model-predicted values corresponding to different hypothesized sequences. 
     
     
         3 . The method of  claim 1 , further comprising:
 assigning a first frequency to the variant sequence;   assigning a second frequency to a non-variant sequence; and   calculating a likelihood of observing each of the variant sequence and the non-variant sequence in the ensemble of sequencing reads based on the first and second frequencies.   
     
     
         4 . The method of  claim 3 , further comprising assigning a third frequency to an outlier event, wherein the outlier event is an event in which a sequence other than the variant sequence and the non-variant sequence occurs. 
     
     
         5 . The method of  claim 4 , wherein the outlier event has a flat density across all sequencing reads in the ensemble. 
     
     
         6 . The method of  claim 4 , wherein the calculating the likelihood further comprises calculating the likelihood of observing an outlier event in the ensemble based further on the third frequency. 
     
     
         7 . The method of  claim 1 , wherein the measurement confidence values are estimated using a function comprising a sum of a log of the likelihood of values measured for a given flow given a hypothesized sequence of the different hypothesized sequences. 
     
     
         8 . The method of  claim 7 , wherein the measurement confidence values are calculated using a function comprising differences between the measured values and the model-predicted values. 
     
     
         9 . The method of  claim 8 , wherein the differences between measured and model-predicted values at each nucleotide flow are modeled to follow independent normal distributions each having a mean and a variance. 
     
     
         10 . The method of  claim 9 , wherein the differences between measured and model-predicted values at each nucleotide flow are modeled to follow independent t-distributions. 
     
     
         11 . The method of  claim 9 , wherein the variance is estimated by decomposition of the variance in a flow and sequencing read into underlying latent components. 
     
     
         12 . The method of  claim 11 , wherein each latent component corresponds to a homopolymer having an integer length. 
     
     
         13 . The method of  claim 11 , wherein the latent components include a null variance component representing contribution to a flow regardless of any nucleotide incorporation, a residual variance component representing contribution for nucleotide incorporations not explicitly modeled, and one or more additional variance components. 
     
     
         14 . The method of  claim 13 , wherein the one or more additional variance components comprise variance components associated with homopolymers having an integer length. 
     
     
         15 . The method of  claim 11 , wherein the latent components are estimated using an expectation-maximization (EM) methodology and a method of moments approximation. 
     
     
         16 . The method of  claim 1 , wherein the measurement confidence values are calculated using an expression comprising
           ∈   i     =     1     1   +   e   x   p       L     L     y   i       −   L     L     x   i                       where LL   yi  and LL xi  are log-likelihoods of values measured for a given sequencing read under hypothesized sequences y and x, respectively. 
     
     
         17 . The method of  claim 16 , wherein the measurement confidence values are calculated using an expression for responsibility comprising
           ρ   i     =     π     π   +       1   −   π       ∗   e   x   p       L     L     y   i       −   L     L     x   i                       where π represents a first frequency assigned to a variant sequence, 1 - π represents a second frequency assigned to a non-variant sequence, and ρ   i  represents a measure of responsibility for each of the sequencing reads in the ensemble. 
     
     
         18 . The method of  claim 1 , further comprising estimating (i) a first frequency π assigned to a variant sequence, (ii) at least one of a measurement confidence value ∈ i  for each of the sequencing reads in the ensemble and a measure of responsibility ρ i  for each of the sequencing reads in the ensemble, and (iii) a variance 
           σ     i   j     2           
 for each of the flows and sequencing reads in the ensemble. 
 
     
     
         19 . A non-transitory machine-readable storage medium comprising instructions which, when executed by a processor, to cause the processor to perform a method comprising:
 receiving an ensemble of sequencing reads based on measurements obtained from a plurality of microwells of a sensor array, the measurements indicative of incorporation of nucleotide species into a plurality of template polynucleotide strands disposed in the plurality of microwells of the sensor array;   assigning measured values to the ensemble of sequencing reads;   calculating model-predicted values for the plurality of template polynucleotide strands utilizing a predictive model of nucleotide incorporations resulting from flows of nucleotide species according to a predetermined order;   modifying at least some model-predicted values using a first bias for forward strands and a second bias for reverse strands, wherein the modifying is based on variations between model-predicted values for different hypothesized sequences obtained using the predictive model of nucleotide incorporations resulting from the flows of nucleotide species according to the predetermined order;   calculating a measurement confidence value for each read in the ensemble of sequencing reads, wherein the measurement confidence value represents variations between the measured values and the modified model-predicted values; and   identifying a plurality of reads in the ensemble as corresponding to a variant sequence based on the measurement confidence value.   
     
     
         20 . A system for identifying sequence variants, including:
 a microwell array comprising a plurality of microwells comprising a plurality of template polynucleotide strands, sequencing primers, and polymerase;   a sensor array comprising a plurality of sensors aligned with the plurality of microwells;   a machine-readable memory operably coupled to the sensor array; and   a processor operably coupled to the sensor array and the machine-readable memory, the processor being configured to execute machine-readable instructions, which, when executed by the processor, causes the system to perform a method comprising: 
 receiving an ensemble of sequencing reads based on measurements obtained from the plurality of sensors of the sensor array, the measurements indicative of incorporation of nucleotide species into a plurality of template polynucleotide strands disposed in the plurality of microwells; 
 assigning measured values to the ensemble of sequencing reads; 
 calculating model-predicted values for the plurality of template polynucleotide strands utilizing a predictive model of nucleotide incorporations resulting from flows of nucleotide species according to a predetermined order; 
 modifying at least some model-predicted values using a first bias for forward strands and a second bias for reverse strands, wherein the modifying is based on variations between model-predicted values for different hypothesized sequences obtained using the predictive model of nucleotide incorporations resulting from the flows of nucleotide species according to the predetermined order; 
 calculating a measurement confidence value for each read in the ensemble of sequencing reads, wherein the measurement confidence value represents variations between the measured values and the modified model-predicted values; and 
 identifying a plurality of reads in the ensemble as corresponding to a variant sequence based on the measurement confidence value.