Integrating variant calls from multiple sequencing pipelines using a machine learning architecture

Abstract

The present disclosure describes methods, non-transitory computer-readable media, and systems that can generate genotype calls from a combined pipeline for processing nucleotide reads from multiple read types / sources for robust and accurate genotype calling. For example, the disclosed systems can train and / or utilize a genotype call integration machine learning model to generate predictions of genotype calls based on data associated with a first type of nucleotide read (e.g., short reads) and a second type of nucleotide read (e.g., long reads). As disclosed, the disclosed systems can determine sequencing metrics and utilize a genotype call integration machine learning model to generate predictions (e.g., genotype probabilities, variant call classifications) for generating output genotype calls based on the sequencing metrics. The disclosed systems can utilize multiple such genotype call integration machine learning models to generate genotype calls for different variants, such as SNPs (Single Nucleotide Polymorphisms) and indels, with the genotype call integration machine learning model generating different predictions for each variant.

Description

[Technical Field]

[0001] (CROSS-REFERENCE TO RELATED APPLICATIONS) This application claims the benefit of and priority to U.S. Provisional Patent Application No. 63 / 482,163, entitled "INTEGRATING VARIANT CALLS FROM MULTIPLE SEQUENCING PIPELINES UTILIZING A MACHINE LEARNING ARCHITECTURE," filed January 30, 2023, and U.S. Provisional Patent Application No. 63 / 378,474, entitled "INTEGRATING VARIANT CALLS FROM MULTIPLE SEQUENCING PIPELINES UTILIZING A MACHINE LEARNING ARCHITECTURE," filed October 5, 2022. The foregoing applications are incorporated herein by reference in their entireties. [Background technology]

[0002] In recent years, biotechnology companies and research institutions have improved the hardware and software for sequencing nucleotides and determining nucleotide base calls for reads, and therefore mutation and genotype calls for genomic samples. For example, some existing nucleosequencing platforms determine individual nucleotide bases (or "nucleobases") within a sequence by using traditional Sanger sequencing or by using sequencing-by-synthesis (SBS) methods. When using SBS, existing platforms can monitor thousands of nucleic acid polymers synthesized in parallel to predict genotype calls from larger base call datasets. For example, cameras in many SBS platforms capture images of illuminated fluorescent tags incorporated into oligonucleotides to determine nucleobase calls. After capturing these images, existing SBS platforms transmit the base call data (or image data) to a computing device to apply sequencing data analysis software that determines the nucleobase sequence of the nucleic acid polymer. Based on the differences between the aligned nucleotide reads and the reference genome, existing systems can further utilize mutation callers to identify mutations in a genomic sample, such as single nucleotide polymorphisms (SNPs), insertions and deletions (indels), and / or structural variations, and genotype calls.

[0003] Despite these recent advances in sequencing and variant calling, existing sequencing systems often contain variant callers that inaccurately call variants, particularly for SNPs and indels. For example, many existing systems generate variant calls that contain an excessive number of false-positive and / or false-negative calls for SNPs and indels. Contributing to this inaccuracy, limitations of some existing sequencing systems dictate that they generate variant calls from a single-stream processing pipeline that focuses on one read source at a time. For example, as alluded to above, some existing systems perform variant calling and / or variant call filtering based solely on nucleotide reads from SBS sequencing. As a further example, some existing systems perform variant calling based solely on nucleotide reads from specific types of long reads, such as circular consensus sequencing (CCS) reads or nanopore long reads. As a result, exclusive reliance on a single source for read data results in many existing systems generating variant calls that contain an excessive number of false positive and / or false negative calls for certain clinical benchmarks that could otherwise be reduced with more accurate systems. Further complicating the issue is that different sequencing systems exhibit different error profiles, such as when conventional systems generate variant calls with higher indel errors based on CCS reads and nanopore long reads compared to sequencing systems that use other types of reads.

[0004] Further complicating this inaccuracy in variant calling is the fact that some existing sequencing systems utilize models that require training on millions or billions of base call data that are either unavailable or incomplete. More specifically, some existing sequencing systems utilize deep learning models that require excessive amounts of training data to achieve acceptable accuracy measures. However, training data for variants is relatively limited for certain variant types (e.g., structural variants), and training models using incomplete or unsubstantial data results in inaccurate and unreliable variant call predictions. Therefore, some existing systems that rely on deep learning models may generate inaccurate variant calls, including SNPs and indels.

[0005] In addition to making inaccurate variant calls, some existing sequencing systems also use overly complex models, wasting computational resources inefficiently. Specifically, the variant callers of some existing sequencing systems are computationally expensive and slow. Indeed, some existing sequencing systems utilize variant callers with deep learning architectures, but training and applying the deep learning architectures requires extensive computational resources (e.g., computational time, processing power, memory). For example, some existing sequencing systems consume hundreds of hours and multiple graphical processing units (GPUs) to train complex convolutional neural networks or other deep learning architectures, which, even after training, consume significant time (e.g., up to 24 hours) across multiple computing devices to generate variant or genotype calls for a single sample sequence.

[0006] A further drawback of existing sequencing systems with complex deep learning networks is that many such systems utilize model architectures that make the sequence data uninterpretable. More specifically, as a basis for generating variant calls, some existing deep neural networks transform and manipulate the sequence data multiple times, changing it from one uninterpretable latent vector to another across various layers and neurons during processing. In many cases, the internal data of these deep neural networks is uninterpretable and cannot be used in any way outside of the neural network architecture itself. Summary of the Invention [Means for solving the problem]

[0007] The present disclosure describes embodiments of a method, a non-transitory computer-readable medium, and a system that can utilize machine learning models to generate predictions of genotype calls based on data from different types of nucleotide reads. In particular, the disclosed system can generate genotype calls from a combined pipeline for processing nucleotide reads from multiple read types / sources for robust and accurate genotype calls (including constituent variant calls). For example, the disclosed system can train or utilize a genotype call integration machine learning model to generate predictions for genotype calls based on data associated with a first type of nucleotide read (e.g., short reads) and a second type of nucleotide read (e.g., long reads). As disclosed, the system can determine sequencing metrics for a first genotype call corresponding to the first type of nucleotide read and a second genotype call corresponding to the second type of nucleotide read. Based on the different or shared sequencing metrics corresponding to the first and second genotype calls, the disclosed systems utilize a genotype call integration machine learning model to generate predictions (e.g., genotype probabilities, variant call classifications) to update or confirm the first or second genotype call, or to determine a different genotype call. In some cases, the disclosed systems may utilize multiple such genotype call integration machine learning models to update or confirm genotype calls for different variants, such as SNPs and indels, but the genotype call integration machine learning models generate different predictions for each variant.

[0008] The detailed description refers to the drawings, which are briefly described below. [Brief explanation of the drawings]

[0009] [Figure 1] FIG. 1 illustrates a block diagram of an exemplary computing environment for implementing a sequencing system and a call integration system, according to one or more embodiments. [Figure 2] 1 illustrates an overview of a call integration system that utilizes a genotype call integration machine learning model to generate genotype calls, according to one or more embodiments. [Figure 3] 1 illustrates exemplary types of nucleotide reads for which a call integration system may generate genotype calls according to one or more embodiments. [Figure 4A] 1 illustrates a call integration system that determines shared or different sequencing metrics between different types of nucleotide reads, according to one or more embodiments. [Figure 4B] 1 illustrates a call integration system that determines shared or different sequencing metrics between different types of nucleotide reads, according to one or more embodiments. [Figure 4C] 1 illustrates a call integration system that determines shared or different sequencing metrics between different types of nucleotide reads, according to one or more embodiments. [Figure 5A] 1 illustrates a call integration system that utilizes a genotype call integration machine learning model to generate predictions (e.g., genotype probabilities or variant call classifications) and corresponding genotype calls, according to one or more embodiments. [Figure 5B] 1 illustrates a call integration system that utilizes a genotype call integration machine learning model to generate predictions (e.g., genotype probabilities or variant call classifications) and corresponding genotype calls, according to one or more embodiments. [Figure 5C] 1 illustrates a call integration system that utilizes a genotype call integration machine learning model to generate predictions (e.g., genotype probabilities or variant call classifications) and corresponding genotype calls, according to one or more embodiments. [Figure 6] FIG. 1 illustrates an exemplary diagram of a training process for learning parameters of a genotype call integration machine learning model, according to one or more embodiments. [Figure 7]FIG. 1 illustrates an exemplary diagram for updating or generating a merged variant call file based on predictions of a genotype call integration machine learning model, according to one or more embodiments. [Figure 8] 1 illustrates example graphs and tables of accuracy metrics for a call integration system in accordance with one or more embodiments. [Figure 9A] 1 illustrates an example table of accuracy metrics for a call integration system according to one or more embodiments. [Figure 9B] 1 illustrates an example table of accuracy metrics for a call integration system according to one or more embodiments. [Figure 10A] 1 illustrates a graph showing accuracy metrics associated with a call integration system in accordance with one or more embodiments. [Figure 10B] 1 illustrates a graph showing accuracy metrics associated with a call integration system in accordance with one or more embodiments. [Figure 11] 1 illustrates a flowchart of a series of operations for generating a genotype call from nucleotide reads of a first read type and a second read type using a genotype call integrated machine learning model, according to one or more embodiments. [Figure 12] 1 illustrates a block diagram of an exemplary computing device for implementing one or more embodiments of the present disclosure. DETAILED DESCRIPTION OF THE INVENTION

[0010] The present disclosure describes embodiments of a call integration system that utilizes a genotype call integration machine learning model to generate and revise genotype calls for genomic samples. In particular, the call integration system may utilize the genotype call integration machine learning model to generate output genotype calls (e.g., reported genotype calls from a merged variant call file) from multiple initial genotype calls (e.g., variant calls) for genomic loci generated by the call generation model from different read types. To generate the output genotype calls, in certain embodiments, the call integration system generates or receives initial genotype calls from read data associated with a combination of short reads (e.g., sequencing-by-synthesis or "SBS" reads) and long reads (e.g., nanopore long reads, circular consensus sequencing or "CCS" reads, and / or assembled nucleotide reads). In some cases, the call integration system determines or identifies specific sequencing metrics (e.g., from read data, call generation model data, and / or external data) to input into the genotype call integration machine learning model to generate the output genotype calls. The call integration system may further train or apply a genotype call integration machine learning model according to the sequencing metrics to generate (or refine or recalibrate) the genotype calls.

[0011] As just mentioned, in certain implementations, the call integration system uses read data from different read types to improve genotype calling accuracy (and corresponding variant calling accuracy). To facilitate generation of genotype calls from multiple read types, in some embodiments, the call integration system receives initial genotype calls from a call generation model. For example, the call integration system (i) receives or determines initial genotype calls (e.g., calls indicating genotypes at genomic coordinates of a nucleotide sequence) corresponding to a first type of nucleotide read (e.g., short reads), and (ii) receives or determines another initial genotype call corresponding to a second type of nucleotide read (e.g., long reads). In some cases, the first type of nucleotide read includes nucleotide reads synthesized from sample library fragments shorter than a first threshold number of nucleobases. Conversely, in the same or other cases, the second type of nucleotide read includes (i) assembled nucleotide reads assembled from shorter nucleotide reads to form a contiguous sequence that meets a second threshold number of nucleobases, (ii) CCS reads that meet the second threshold number of nucleobases, and / or (iii) nanopore long reads that meet the second threshold number of nucleobases.

[0012] From the initial genotype calls corresponding to different read types, the call integration system may further generate output genotype calls, such as the presence or absence of variants such as SNPs or indels, and predictions of allele zygosity for the genomic sample. As mentioned, to generate the output genotype calls, the call integration system may extract, identify, or determine sequencing metrics (associated with the initial genotype calls from different read types) and input them into a genotype call integration machine learning model. The genotype call integration machine learning model then generates a set of likelihoods or predictions indicating the likelihood that the initial genotype call is correct or incorrect (e.g., a different set of predictions for each initial genotype call corresponding to a different read type and / or for each different variant type). For example, the call integration system may extract or determine sequencing metrics belonging to one or more categories, including (i) read-based sequencing metrics, (ii) call model-generated sequencing metrics, and (iii) externally sourced sequencing metrics. Additional details regarding the configuration and determination of sequencing metrics are provided below with reference to the figures.

[0013] As suggested, in certain embodiments, the call integration system generates genotype calls using a multi-stream pipeline that processes multiple read types for output genotype calls as part of a combined or merged variant call file based on the multiple read types. For example, the call integration system (i) processes a first set of sequencing metrics extracted from the initial genotype calls based on a first read type, and (ii) processes a second set of sequencing metrics extracted from the initial genotype calls based on a second read type. Additionally, the call integration system may utilize a genotype call integration machine learning model to generate a set of predictions based on the first and second sets of sequencing metrics, and generate the output genotype call from the set of predictions.

[0014] To predict or generate output genotype calls for different variants (e.g., SNPs and indels), in some cases, the call integration system utilizes a genotype call integration machine learning model to generate different prediction sets for different variants (e.g., from the same sequencing metrics or different sequencing metrics). For example, the call integration system may utilize a first instance of a genotype call integration machine learning model (e.g., trained to predict SNPs) to process first and second read-based sequencing metrics (e.g., SBS sequencing metrics) to generate an output genotype call for a SNP at a genomic coordinate. Additionally, the call integration system may utilize a second genotype call integration machine learning model (e.g., trained to predict indels) to process the first and second read-based sequencing metrics to generate an output genotype call for an indel at a different (or the same) genomic coordinate. In some embodiments, the call integration system may utilize a first genotype call integration machine learning model for biallelic SNPs, but a second genotype call integration machine learning model for other types of variant calls (e.g., variants that are not biallelic SNPs). Furthermore, although this disclosure describes at least two different types of genotype call integration machine learning models, in certain implementations, the call integration system trains or applies a single genotype call integration machine learning model to generate both genotype predictions for different types of variants (e.g., genotype predictions for either SNPs or indels).

[0015] As alluded to above, the call integration system provides several advantages, benefits, and / or improvements over existing sequencing systems, including mutation callers and other sequencing data analysis software. For example, the call integration system generates more accurate genotype calls (including mutation calls) than existing sequencing systems. While some conventional sequencing systems generate inaccurate variant calls (particularly for SNPs and indels), the call integration system trains or utilizes a genotype call integration machine learning model to improve genotype / variation calling over conventional systems. Specifically, unlike conventional systems that rely on a single source for read data, the call integration system can process multiple reads of different types (e.g., assembled nucleotide reads and SBS reads) to generate more accurate genotype calls corresponding to SNPs and indels (thereby reducing false positives and false negatives). Also contributing to improved accuracy over conventional systems, the call integration system can utilize different instances of a genotype call integration machine learning model trained on different variant types (e.g., SNPs and indels) to generate or predict genotype calls from multiple read types, which conventional systems cannot do. Further contributing to improved accuracy in genotype calling, in some cases, the call integration system determines and utilizes specific sequencing metrics (inherent in conventional systems) as criteria (e.g., as input data) for generating calls via the genotype call integration machine learning model.

[0016] To achieve the aforementioned accuracy improvements, the call integration system utilizes an improved, unique machine learning model, i.e., a genotype call integration machine learning model, that is trained to perform the new application. Unlike existing mutation callers that generate genotype calls from general single-stream sequencing data (without tuning or emphasis on whether a particular genomic coordinate has historically indicated or been detected to indicate a particular mutation), the call integration system utilizes (multiple instances of) a unique genotype call integration machine learning model that generates specific predictions or classifications for different types of mutations (e.g., SNPs and indels) from multi-read data. In some cases, the call integration system utilizes the genotype call integration machine learning model as a post-processing filter to either (i) select between a first genotype call corresponding to a first type of nucleotide read and a second genotype call corresponding to a second type of nucleotide read, or (ii) determine another genotype call that is different from the first and second genotype calls.

[0017] Attributing at least in part to the improved accuracy, the call integration system exhibits improved flexibility over existing sequencing systems. For example, while many existing sequencing systems are limited to analyzing read data from one read type at a time, in some embodiments, the call integration system is adapted to process multiple read types to merge data and generate an output genotype call for a particular genomic coordinate or region. Specifically, unlike some existing sequencing systems, the call integration system can generate a genotype call (e.g., including a variant call) for a genomic coordinate based on multiple types of read data for the genomic coordinate, such as assembled nucleotide reads and SBS reads.

[0018] In addition to improving accuracy and flexibility, in certain embodiments, the call integration system improves computational efficiency and speed. As described above, some existing sequencing systems utilize computationally expensive and slow neural network architectures (e.g., deep learning architectures such as convolutional neural networks) that require many hours (e.g., up to 24 hours) across multiple high-end processing devices to process read data and generate calls for genomic samples. Furthermore, the call integration system can generate a (merged) variant call file by updating only certain fields without regenerating an entirely new variant call file (as is done by some conventional systems). Such deep learning architectures may also require several days (or weeks) to train. Conversely, the call integration system utilizes a relatively lightweight and fast architecture for the genotype call integration machine learning model. In contrast to the many hours across multiple processing devices required by existing sequencing systems, the call integration system requires a runtime (e.g., on a single processing device) of less than one hour (e.g., approximately 15 minutes for the call generation model and less than one minute for the genotype call integration machine learning model) to generate genotype calls (and / or variant calls) for a genomic sample. Therefore, the call integration system is much faster and less computationally expensive than many deep learning approaches to genotype calling / variant calling. In fact, not only are the call integration system's models fast and computationally inexpensive to implement, but the genotype call integration machine learning models are also much faster and less computationally expensive than many existing deep learning systems.

[0019] As an additional advantage over existing sequencing systems, in certain implementations, the call integration system can identify or facilitate changes in individual sequencing metrics that affect the accuracy of genotype calls (and corresponding variant calls). While the neural network architecture of many existing sequencing systems makes interpretation of internal model data impossible due to latent features hidden among their many layers and neurons, the call integration system utilizes a model architecture that facilitates interpretation of the impact of individual sequencing metrics. More specifically, in some cases, the call integration system utilizes a call generation model and a genotype call integration machine learning model to much more easily extract and analyze the individual sequencing metrics used throughout the process of generating a genotype call. Indeed, the call integration system can determine the contribution measure of each of the sequencing metrics involved in determining a genotype call in a particular region of genomic coordinates.

[0020] As suggested by the foregoing discussion, the present disclosure utilizes various terms to describe the features and advantages of the call integration system. Further details regarding the meaning of these terms as used in this disclosure are provided below. As used in this disclosure, for example, the terms "sample nucleotide sequence" or "sample sequence" refer to a sequence of nucleotides isolated or extracted from a sample organism (or a copy of such an isolated or extracted sequence). In particular, a sample nucleotide sequence is isolated or extracted from a sample organism and includes a segment of a nucleic acid polymer composed of nitrogenous heterocyclic bases. For example, a sample nucleotide sequence may include a segment of deoxyribonucleic acid (DNA), ribonucleic acid (RNA), or other polymeric forms of nucleic acid or chimeric or hybrid forms of nucleic acids as described below. More specifically, in some cases, a sample nucleotide sequence is one found in a sample prepared or isolated by a kit and received by a sequencing instrument.

[0021] Relatedly, as used herein, the term "genomic sample" refers to a target genome or portion of a genome that is to be assayed or sequenced. For example, a genomic sample includes one or more sequences of nucleotides isolated or extracted from a sample organism (or copies of such isolated or extracted sequences). In particular, a genomic sample includes the entire genome, isolated or extracted (in whole or in part) from a sample organism and composed of nitrogenous heterocyclic bases. For example, a genomic sample may include segments of deoxyribonucleic acid (DNA), ribonucleic acid (RNA), or other polymeric forms of nucleic acid or chimeric or hybrid forms of nucleic acids as described below. In some cases, a genomic sample is one found in a sample prepared or isolated by a kit and received by a sequencing instrument.

[0022] As further used herein, the term "genotype call" refers to the determination or prediction of a specific genotype of a genomic sample at a genomic locus. In particular, a genotype call can include the prediction of a specific genotype of a genomic sample relative to a reference genome or reference sequence at a genomic coordinate or genomic region. For example, in some cases, a genotype call includes the determination or prediction that a genomic sample contains both a nucleobase and a complementary nucleobase at a genomic coordinate that is either homozygous or heterozygous for the reference base or mutation (e.g., a homozygous reference base represented as 0|0, or heterozygous for a mutation on a particular strand represented as 0|1). Thus, a genotype call can include the prediction of a mutation or reference base for one or more alleles of a genomic sample and can indicate the zygosity for the mutation or reference base. Genotype calls are often determined for genomic coordinates or genomic regions where SNPs, insertions, deletions, or other mutations have been identified for a population of organisms.

[0023] In some cases, an "initial genotype call" refers to a genotype call corresponding to or determined from nucleotide read data and / or sequencing metrics for a particular type of nucleotide read. For example, the initial genotype call may include a first genotype call corresponding to a first type of nucleotide read of a first threshold number of nucleobases and / or a second genotype call corresponding to a second type of nucleotide read of a second threshold number of nucleobases. In contrast, an "output genotype call" refers to a genotype call reported by or generated for an output data file. For example, the output genotype call may include a final genotype call determined based on one or both of genotype probabilities and variant call classifications from a genotype call integrated machine learning model and included in a variant call file (VCF).

[0024] As further used herein, the term "nucleobase call" (or simply "base call") refers to the determination or prediction of a particular nucleobase (or nucleobase pair) for an oligonucleotide (e.g., a nucleotide read) during a sequencing cycle or for a genomic coordinate of a sample genome. In particular, nucleobase calling can refer to (i) the determination or prediction of the type of nucleobase incorporated within an oligonucleotide on a nucleotide-sample slide (e.g., a read-based nucleobase call), or (ii) the determination or prediction of the type of nucleobase present at a genomic coordinate or region within a genome, including a variant or non-variant call in a digital output file. In some cases, for a nucleotide read, nucleobase calling includes the determination or prediction of a nucleobase based on intensity values ​​obtained from fluorescently tagged nucleotides attached to oligonucleotides on a nucleotide-sample slide (e.g., in a cluster of a flow cell). Alternatively, nucleobase calling includes the determination or prediction of a nucleobase from a chromatogram peak or current change resulting from a nucleotide passing through a nanopore on a nucleotide-sample slide. In contrast, a nucleobase call may also include a final prediction of a nucleobase at a genomic coordinate of a sample genome for a variant call file (VCF) or another base call output file based on the nucleotide reads corresponding to the genomic coordinate. Thus, a nucleobase call may include a base call corresponding to a genomic coordinate and a reference genome, such as an indication of a mutation or non-mutation at a specific position corresponding to the reference genome. In practice, a nucleobase call may refer to a variant call, including but not limited to a single nucleotide variant (SNV), an insertion or deletion (indel), or a base call that is part of a structural variant. As suggested above, a single nucleobase call may be an adenine (A) call, a cytosine (C) call, a guanine (G) call, a thymine (T) call, or a uracil (U) call.

[0025] Relatedly, as used herein, the term "nucleotide read" refers to a deduced sequence of one or more nucleotide bases (or nucleobase pairs) from all or a portion of a sample's nucleotide sequence (e.g., a sample's genomic sequence, complementary DNA). In particular, a nucleotide read includes a determined or predicted nucleobase call for a nucleotide fragment (or group of monoclonal nucleotide fragments) from a sequencing library corresponding to a genomic sample. For example, in some embodiments, a call integration system determines a nucleotide read by generating a nucleobase call for nucleobases that have passed through a nanopore in a nucleotide-sample slide, or via fluorescent tagging, or from a well in a flow cell. In some cases, a nucleotide read may refer to a specific type of read, such as a nucleotide read (e.g., an SBS read) synthesized from a sample library fragment shorter than a threshold number of nucleobases. In these or other cases, another type of nucleotide read may refer to (i) a nucleotide read assembled from shorter nucleotide reads to form a continuous sequence (e.g., an assembled nucleotide read) that meets a threshold number of nucleobases, (ii) a circular consensus sequencing (CCS) read that meets a threshold number of nucleobases, or (iii) a nanopore long read that meets a threshold number of nucleobases.

[0026] As described above, in some embodiments, the call integration system determines sequencing metrics for the nucleobase calls of the nucleotide reads. As used herein, the term "sequencing metric" refers to a quantitative measure or score that indicates the degree to which individual nucleobase calls (or a sequence of nucleobase calls) align, compare, or quantify with respect to genomic coordinates or genomic regions of a reference genome, with respect to nucleobase calls from nucleotide reads, or with respect to external genome sequencing or genomic structure. For example, sequencing metrics include quantitative measures or scores that indicate (i) the degree to which individual nucleobase calls align, map, or cover genomic coordinates or reference bases of a reference genome, (ii) the degree to which nucleobase calls compare with reference or alternative nucleotide reads with respect to mapping, mismatches, base call quality, or other raw sequencing metrics, or (iii) the degree to which genomic coordinates or regions corresponding to nucleobase calls demonstrate mappability, repeat base call content, DNA structure, or other generalized metrics.

[0027] Along these lines, the call integration system determines various types of sequencing metrics from different sources, such as read-based sequencing metrics, external source sequencing metrics, and call model generation sequencing metrics. As used herein, the term "read-based sequencing metrics" refers to sequencing metrics derived from nucleotide reads of a sample nucleotide sequence. For example, read-based sequencing metrics include sequencing metrics determined by applying statistical tests to detect differences between a reference sequence and a nucleotide read. In some embodiments, read-based sequencing metrics may include comparative mapping quality distribution metrics indicating a comparison between mapping qualities or comparative mismatch count metrics indicating a comparison between mismatch counts. In some cases, read-based sequencing metrics may correspond to nucleobase calls generated from different read types, such as assembled nucleotide reads and / or SBS reads.

[0028] In contrast, "externally sourced sequencing metrics" refer to sequencing metrics identified or obtained from one or more external databases. For example, externally sourced sequencing metrics include metrics related to nucleotide mappability, replication timing, or DNA structure that are available outside the call integration system.

[0029] Furthermore, the term "call-model-generated sequencing metrics" refers to internal, model-specific sequencing metrics generated or extracted by a call-generation model. For example, call-model-generated sequencing metrics include variant calling sequencing metrics extracted or determined via the variant calling component of the call-generation model and mapping and alignment sequencing metrics extracted or determined via the mapping and alignment component of the call-generation model. As noted above, call-model-generated sequencing metrics can include alignment metrics, such as deletion size metrics or mapping quality metrics, that quantify the degree to which a sample nucleic acid sequence aligns with the genomic coordinates of an exemplary nucleic acid sequence. Furthermore, call-model-generated sequencing metrics can include depth metrics, such as forward-reverse depth metrics or normalized depth metrics, that quantify the depth of nucleobase calling for a sample nucleic acid sequence at the genomic coordinates of an exemplary nucleic acid sequence. The sequencing metrics for call model generation may also include a call quality metric that quantifies the quality or accuracy of a nucleobase call, such as a nucleobase-call-quality metric, a callability metric, or a somatic-quality metric.

[0030] As further used herein, the term "genomic coordinate" (or sometimes simply "coordinate") refers to a specific location or position of a nucleobase within a genome (e.g., the genome of an organism or a reference genome). In some cases, a genomic coordinate includes an identifier for a specific chromosome of a genome and an identifier for the location of a nucleobase within the specific chromosome. For example, a genomic coordinate(s) may include a chromosome (e.g., chr1 or chrX) number, name, or other identifier, and a specific location(s), such as a numbered location following the identifier of the chromosome (e.g., chr1:1234570 or chr1:1234570-1234870). In some cases, a genomic coordinate refers to a genomic coordinate on a sex chromosome (e.g., chrX or chrY). As a result, the call integration system may determine a genotype probability and / or a variant call classification for a genotype call (e.g., a variant call) for a genomic coordinate on a sex chromosome. Furthermore, in certain implementations, a genomic coordinate refers to the source of the reference genome (e.g., mt for a mitochondrial DNA reference genome, or SARS-CoV-2 for the reference genome of the SARS-CoV-2 virus) and the position of a nucleobase within the source for the reference genome (e.g., mt:16568 or SARS-CoV-2:29001). In contrast, in certain cases, a genomic coordinate refers to the position of a nucleobase within a reference genome without reference to a chromosome or source (e.g., 29727).

[0031] Also, as used herein, the term "genomic region" refers to a range of genomic coordinates. Similar to genomic coordinates, in certain embodiments, a genomic region can be identified by a chromosomal identifier and a specific location(s), such as a numbered location following the chromosomal identifier (e.g., chr1:1234570-1234870).

[0032] As noted above, genome coordinates include locations within a reference genome. Such locations may be within a particular reference genome. As used herein, the term "reference genome" refers to a digital nucleic acid sequence assembled as a representative example (or multiple representative examples) of an organism's genes and other genetic sequences. Regardless of sequence length, in some cases, a reference genome represents an exemplary set of genes or a set of nucleic acid sequences in a digital nucleic acid sequence determined by scientists as representative of a particular species of organism. For example, a linear human reference genome may be GRCh38 or other versions of the reference genome from the Genome Reference Consortium. As a further example, a reference genome may include a reference graph genome, e.g., Illumina DRAGEN Graph Reference Genome hg19, that includes both a linear reference genome and paths representing nucleic acid sequences from ancestral haplotypes.

[0033] Furthermore, as used herein, the term "reference multigenome" (sometimes referred to as "graph reference genome") refers to a reference genome that includes both a linear reference genome and alternative contiguous sequences (or graph extensions) that represent variant haplotype sequences or other variant or alternative nucleic acid sequences. For example, a reference multigenome can include a linear reference genome and alternative contiguous sequences that correspond to one or more population haplotype sequences identified from a genome sample database. As an example, a reference multigenome can include Illumina DRAGEN Graph Reference Genome hg19.

[0034] As further used herein, the term "contiguous sequence" (or "contig assembly") refers to a consensus nucleotide sequence for a genomic region of a genomic sample (or multiple genomic samples of a species) based on a set of overlapping nucleotide segments corresponding to the genomic region. In particular, a contiguous sequence comprises a consensus nucleotide sequence for a genomic region of one or more genomic samples based on nucleotide reads for the one or more genomic samples that cover (or overlap with) the genomic region. As noted above, the terms "contiguous sequence" and "contig assembly" may be used interchangeably.

[0035] Relatedly, the term "alternate contiguous sequence" (or simply "alt contig") refers to a contiguous sequence representing a population haplotype that has been added to a linear reference genome (or other reference genome) at a particular genomic coordinate(s) (e.g., lifted into the linear reference genome). In some implementations, a graph reference genome (or reference multi-genome) may include alternative contiguous sequences mapped to genomic coordinates of the primary assembly for the linear reference genome. For example, an alternative contiguous sequence may represent a population haplotype containing a mutation with liftover to two or more genomic coordinates in the linear reference genome corresponding to two or more sides of a variant interchromosomal translocation. In some cases, the hash table of the graph reference genome (or reference multi-genome) includes identifiers that associate alternative contiguous sequences representing variant haplotypes with genomic coordinates representing reference haplotypes from the primary assembly of the linear reference genome.

[0036] As used herein, the term "base-call-quality metric" refers to a specific score or other measure that indicates the accuracy of a nucleobase call. In particular, a base-call quality metric includes a value that indicates the likelihood that one or more predicted nucleobase calls for a genomic coordinate will contain an error. For example, in certain implementations, a base-call quality metric may include a Q-score (e.g., a PHil's Read EDitor (PHRED) quality score) that predicts the probability of error for any given nucleobase call. Illustratively, a quality score (or Q-score) may indicate that the probability of an incorrect nucleobase call at a genomic coordinate is equal to 1 in 100 for a Q20 score, 1 in 1,000 for a Q30 score, 1 in 10,000 for a Q40 score, etc.

[0037] Relatedly, in some embodiments, the call integration system may generate sequencing metrics through modifying or updating previous metrics. Such “re-engineered sequencing metrics” may refer to sequencing metrics that have been updated, modified, augmented, refined, or re-engineered to measure or compare nucleobase calls (e.g., nucleobase calls for reads, genotypes, or variant calls) against other nucleobase calls, standards, or references, or targeted for a particular purpose or task. For example, re-engineered sequencing metrics may include modifications to raw (e.g., unmodified) sequencing metrics or combinations of raw (e.g., unmodified) sequencing metrics. In some embodiments, for example, the call integration system generates one or more of read-based sequencing metrics, external source sequencing metrics, and / or call model generation sequencing metrics as re-engineered sequencing metrics. In some cases, re-engineered sequencing metrics refer to sequencing metrics that are generated by the call integration system and are therefore proprietary to or internal to the call integration system and are not available to third-party systems. Exemplary re-engineered sequencing metrics include comparative mapping quality distribution metrics that indicate a comparison between mapping quality distributions associated with the reference sequence and the alternative supporting nucleotide read, or comparative base quality metrics that indicate a comparison between the base qualities of the reference sequence and the alternative supporting nucleotide read.

[0038] As alluded to above, the call integration system may utilize machine learning models to modify sequencing metrics and update nucleobase calls. As used herein, the term "machine learning model" refers to a computer algorithm or collection of computer algorithms that automatically improves at a particular task through experience based on the use of data. For example, a machine learning model may utilize one or more learning techniques to improve accuracy and / or effectiveness. Exemplary machine learning models include various types of decision trees (e.g., gradient boosted trees), support vector machines, Bayesian networks, or neural networks.

[0039] In some cases, the call integration system utilizes a genotype call integration machine learning model to generate, modify, or update predictions of genotype calls based on sequencing metrics. As used herein, the term "genotype-call-integration machine-learning model" refers to a machine learning model that generates predictions, such as genotype probabilities and / or variant call classifications, for one or more genomic samples. As indicated above, a genotype call integration machine learning model includes a machine learning model that generates predictions for genotype calls for one or more genomic samples based on data from different types of nucleotide reads. For example, in some cases, a genotype call integration machine learning model is trained to generate genotype probabilities that indicate the probability or likelihood of various genotypes at one or more genomic coordinates based on sequencing metrics. As another example, a genotype call integration machine learning model is trained to generate variant call classifications that indicate various probabilities or predictions of variant calls based on sequencing metrics. In some cases, the genotype call integration machine learning model is a series of gradient-boosted decision trees (e.g., the XGBoost algorithm or a treelite algorithm for an ensemble of decision trees), while in other cases, the genotype call integration machine learning model is a random forest model, a multilayer perceptron, a linear regression, a support vector machine, a deep table learning architecture, a deep learning transformer (e.g., a self-attention-based table transformer), or a logistic regression. In certain embodiments, the genotype call integration machine learning model includes multiple submodels or operates in conjunction with another (instance of the genotype call integration machine learning model). For example, a first genotype call integration machine learning model (e.g., an ensemble of gradient-boosted trees) generates a first set of predictions for a first variant (e.g., SNP) at a genomic coordinate, and a second genotype call integration machine learning model generates a second set of predictions for a second variant (e.g., indel) at a genomic coordinate.

[0040] Relatedly, the term "variant call classification" refers to a predicted classification from a genotype call integrated machine learning model, indicating a probability, score, or other quantitative measure associated with some aspect of a genotype call (and how the genotype call affects the variant call) based on one or more sequencing metrics. Variant call classifications may include specialized predictions depending on the application of the genotype call integrated machine learning model, such as for predicting indels. For example, variant call classifications may include, but are not limited to, (i) a true-positive mutation probability, where the genotype call constitutes a true-positive mutation for one or more genomic coordinates of a genomic sample; (ii) a zygosity error probability, where the genotype call includes a genotype-zygosity error at one or more genomic coordinates; and (iii) a reference probability of a homozygous reference genotype at one or more genomic coordinates. Thus, the term "reference probability" may refer to the probability of a homozygous reference genotype occurring at one or more genomic coordinates. As described below, in some cases, the genotype call integrated machine learning model generates a variant call classification based on a first type of nucleotide read (e.g., SBS read) and a second type of nucleotide read (e.g., assembled nucleotide read).

[0041] As further used herein, the term "genotype probability" refers to the likelihood, probability, or score of a particular genotype at a genomic coordinate or genomic region. For example, genotype probability includes the likelihood of a homozygous reference genotype, the likelihood of a heterozygous variant genotype, or the likelihood of a homozygous variant genotype at one or more genomic coordinates. In some cases, genotype probability may refer to a posterior genotype probability. Thus, in some cases, the genotype probability determined by a genotype call integration machine learning model may be presented (or modified to be presented) in the posterior genotype probability (GP) field of a VCF, such as a merged VCF. The genotype probability may include a specific prediction depending on the application of the genotype call integration machine learning model, such as for predicting SNPs.

[0042] As described above, the call integration system may generate genotype probabilities and / or variant call classifications that indicate or reflect the likelihood of identifying a variant at a genomic coordinate. As used herein, the term "variant" refers to a nucleobase or nucleobases that do not align with, differ from, or are different from the corresponding nucleobase(s) in a reference sequence or reference genome. For example, variants include SNPs, indels, or structural variations that represent nucleobases in a sample's nucleotide sequence that differ from the nucleobases of the corresponding genomic coordinates of the reference sequence.

[0043] As mentioned, in some embodiments, the call consolidation system modifies data fields corresponding to the variant call file. As used herein, the term "variant call file" refers to a digital file that indicates or represents one or more nucleobase calls and / or variant calls compared to a reference genome, along with other information about the calls. In some cases, the variant call file may also include genotype calls for a genomic sample that indicate the reference or variant call for an allele at a specific genomic coordinate or region. For example, a variant call format (VCF) file refers to a text file format that contains information about variants at specific genomic coordinates, including a meta-information row, a header row, and data rows, each data row having information about a single nucleobase call (e.g., a single mutation). As described further below, the call consolidation system may generate different versions of the variant call file, including a pre-filter variant call file that contains variant nucleobase calls that either pass or do not pass a quality filter for a base call quality metric, or a post-filter variant call file that contains variant nucleobase calls that pass a quality filter but exclude those that do not pass the quality filter.

[0044] Relatedly, a "merged variant call file" refers to a variant call file generated from one or more other variant call files. For example, a merged variant call file refers to a variant call file generated by selecting or merging data from a variant call file associated with one or more genotype calls determined from a first type of nucleotide read and a variant call file associated with one or more genotype calls determined from a second type of nucleotide read. In some cases, a merged variant call file includes some data selected from one initial variant call file and other data selected from a different initial variant call file. Note that a merged variant call file may include data from merged positions, where some fields are generated to include new data not found in other (e.g., non-merged) variant call files. Thus, in some embodiments, a merged variant call file is generated from initial variant call files associated with different types of nucleotide calls.

[0045] In some embodiments, the call integration system modifies data fields corresponding to metrics for nucleobase calls associated with a variant call file, such as fields for call quality, genotype, and genotype quality. As used herein, the term "call quality," when used with respect to a data field in a variant call file, refers to a measure or indicator of the likelihood or probability that a variant is present at a given location. Thus, a call quality field (or QUAL field) corresponding to a VCF file may include a base call quality metric, such as a PHRED-scaled quality or Q-score, which represents the probability that a genomic coordinate in a sample genome contains a variant. Similarly, "genotype quality," when used with respect to a field, refers to the likelihood or probability that a particular predicted genotype for a nucleobase call is accurate.

[0046] As described, in some embodiments, the call integration system utilizes a call generation model to generate nucleobase calls for genomic coordinates. As used herein, the term "call generation model" refers to a probabilistic model that generates sequencing data, including nucleobase calls, variant calls, and / or genotype calls, along with associated metrics, from nucleotide reads of a sample nucleotide sequence. Thus, in some cases, the call generation model may be a variant call generation model. For example, in some cases, the call generation model refers to a Bayesian probability model that generates variant calls based on nucleotide reads of a sample nucleotide sequence. Such a model may process or analyze sequencing metrics corresponding to a read pileup (e.g., multiple nucleotide reads corresponding to a single genomic coordinate), including mapping quality, base quality, and various hypotheses including extraneous reads, missing reads, joint detection, etc. The call generation model may also include multiple components, including, but not limited to, different software applications or components for mapping and alignment, sorting, duplicate marking, calculation of read pileup depth, and variant calling. In some cases, the call generation model refers to the ILLUMINA DRAGEN model for variant calling and mapping and alignment functions (e.g., DRAGEN variant caller or "DRAGEN VC").

[0047] The following paragraphs describe the call integration system with reference to example diagrams depicting example embodiments and implementations. For example, Figure 1 illustrates a schematic diagram of a system environment (or "environment") 100 in which a call integration system 106 operates, according to one or more embodiments. As shown, environment 100 includes one or more server device(s) 102 connected to a client device 108, a local device 116, and a sequencing device 114 via a network 112. While Figure 1 illustrates one embodiment of call integration system 106, this disclosure below describes alternative embodiments and configurations.

[0048] 1, server device(s) 102, client device 108, local server device 116, and sequencing device 114 may communicate with each other via network 112. Network 112 includes any suitable network with which computing devices may communicate. An exemplary network is described in further detail below with respect to FIG. 12.

[0049] As illustrated by FIG. 1 , the sequencing device 114 includes a device for sequencing nucleic acid polymers. In some embodiments, the sequencing device 114 analyzes nucleic acid segments or oligonucleotides extracted from a genomic sample to generate nucleotide reads or other data using computer-implemented methods and systems (described herein) either directly or indirectly on the sequencing device 114. More specifically, the sequencing device 114 receives and analyzes nucleic acid sequences extracted from a genomic sample in a nucleotide-sample slide. In one or more embodiments, the sequencing device 114 utilizes SBS to sequence the nucleic acid polymer into nucleotide reads. In some embodiments, the sequencing device 114 communicates directly with the client device 108, bypassing the network 112, in addition to or instead of communicating via the network 112.

[0050] As further illustrated by FIG. 1 , the local device 116 is located at or near the same physical location as the sequencing device 114. Indeed, in some embodiments, the local device 116 and the sequencing device 114 are integrated into the same computing device. The local device 116 may generate, receive, analyze, store, and transmit digital data, such as by executing a call integration system 106 to receive sequencing metrics or to determine genotype and / or variant calls based on analysis of such sequencing metrics. As illustrated in FIG. 1 , the sequencing device 114 may transmit (and the local device 116 may receive) sequencing metrics generated during the sequencing run of the sequencing device 114. By executing software in the form of the call integration system 106, the local device 116 may align nucleotide reads to a reference genome and / or utilize a genotype call integration machine learning model 107 to determine genotypes and / or genetic variants based on the sequencing metrics. The local server device 116 may also communicate with the client device 108. In particular, the local device 116 may transmit data to the client device 108, including a variant call file (VCF), sequencing metrics, or other information indicative of nucleobase calls, genotype calls, variant calls, sequencing metrics, error data, or other metrics.

[0051] As further illustrated by FIG. 1, server device(s) 102 may generate, receive, analyze, store, and transmit digital data, such as data for determining genotype calls or for sequencing nucleic acid polymers. As illustrated in FIG. 1, sequencing device 114 may transmit call data and / or sequencing metrics (and server device(s) 102 and / or local device 116 may receive them). Server device(s) 102 may also communicate with client device 108 and / or local device 116. In particular, server device(s) 102 and / or local device 116 may transmit to client device 108 data including variant call files or other information indicative of nucleobase calls, genotype calls, variant calls, sequencing metrics, error data, or other metrics.

[0052] In some embodiments, server device(s) 102 comprise a distributed collection of servers, where server device(s) 102 include multiple server devices distributed across network 112 and located in the same or different physical locations. Further, server device(s) 102 may comprise content servers, application servers, communication servers, web hosting servers, or another type of server. In some cases, server device(s) 102 are located in the same physical location as sequencer 114 and / or local device 116.

[0053] 1 , the server device(s) 102 and / or the sequencing device 114 may include a sequencing system 104. Generally, the sequencing system 104 analyzes read data and / or call data, such as sequencing metrics, received from the sequencing device 114 to determine a nucleobase sequence for the nucleic acid polymer. For example, the sequencing system 104 may receive raw data from the sequencing device 114 and determine a nucleobase sequence for the nucleic acid segment. In some embodiments, the sequencing system 104 determines the sequence of nucleobases in a DNA and / or RNA segment or oligonucleotide. In addition to processing and determining a sequence for the nucleic acid polymer, the sequencing system 104 also generates a variant call file indicating one or more genotype calls and / or variant calls for one or more genomic coordinates.

[0054] As described above and illustrated in FIG. 1 , the call integration system 106 analyzes call data, such as sequencing metrics, from the sequencing device 114 to determine genotype calls for sample nucleotide sequences of a genomic sample. The call integration system 106 includes a call generation model and a genotype call integration machine learning model 107. In some embodiments, the call integration system 106 determines sequencing metrics for the sample nucleotide sequences. Based on data derived or prepared from the sequencing metrics, the call integration system 106 trains and / or applies the call generation model to determine nucleobase calls for the sample sequences corresponding to genomic coordinates. The call integration system 106 further utilizes the genotype call integration machine learning model 107 to generate a set of predictions (e.g., genotype probabilities for SNPs or variant call classifications for indels) to update or revise the genotype calls (and / or variant calls). Based on such data, the call integration system 106 may, for example, update data fields corresponding to variant call files to update the genotype calls and / or variant calls to improve accuracy.

[0055] 1 , client device 108 may generate, store, receive, and transmit digital data. In particular, client device 108 may receive sequencing metrics from sequencing device 114. Additionally, client device 108 may communicate with server device(s) 102 and / or local device 116 to receive variant call files containing genotype calls and / or other metrics, such as call quality and / or genotype quality. Accordingly, client device 108 may present or display information about the genotype calls in a graphical user interface to a user associated with client device 108. For example, client device 108 may present a contribution measure interface that includes a visualization or depiction of various contribution measures associated with or attributable to individual sequencing metrics for a particular nucleobase call.

[0056] 1 may include various types of client devices. For example, in some embodiments, client device 108 includes a non-mobile device such as a desktop computer or server, or other type of client device. In still other embodiments, client device 108 includes a mobile device such as a laptop, tablet, mobile phone, or smartphone. Further details regarding client device 108 are discussed below with respect to FIG. 12.

[0057] 1 , client device 108 includes a sequencing application 110. Sequencing application 110 may be a web application or a native application (e.g., a mobile application, a desktop application) stored and executed on client device 108. Sequencing application 110 may include instructions that (when executed) cause client device 108 to receive data from call integration system 106 and present data from variant call files for display on client device 108. Additionally, sequencing application 110 may instruct client device 108 to display a visualization of the contribution measures for sequencing metrics of genotype calls.

[0058] 1 , the call integration system 106 may be located on the client device 108 as part of the sequencing application 110, or on the sequencing device 114 or the local device 116. Thus, in some embodiments, the call integration system 106 is implemented (e.g., located completely or partially) on the client device 108. In still other embodiments, the call integration system 106 is implemented by one or more other components of the environment 100, such as the sequencing device 114 and the local device 116. In particular, the call integration system 106 may be implemented in a variety of different ways across the server device(s) 102, the network 112, the client device 108, and the sequencing device 114. For example, the call integration system 106 may be downloaded from the server device(s) 102 to the client device 108, the local device 116, and / or the sequencing device 114, with all or a portion of the functionality of the call integration system 106 being implemented on each device in the environment 100.

[0059] 1 illustrates components of environment 100 communicating via network 112, in certain implementations, components of environment 100 may also communicate directly with one another, bypassing network 112. For example, as previously discussed, in some implementations, client device 108 communicates directly with sequencing device 114 and / or local device 116. Note that in some embodiments, client device 108 communicates directly with call integration system 106 (hosted on one or more of the illustrated components). Furthermore, call integration system 106 may access one or more databases housed on or accessed by server device(s) 102 or elsewhere in environment 100.

[0060] As indicated above, the call integration system 106 may determine output genotype calls based on one or more sequencing metrics for initial genotype calls from different types of nucleotide reads. In particular, the call integration system 106 may utilize a genotype call integration machine learning model to generate predictions (e.g., genotype probabilities or variant call classifications) from the sequencing metrics and may determine or update various metrics (e.g., in a VCF file) associated with the genotype calls from the generated predictions. According to one or more embodiments, FIG. 2 illustrates an exemplary overview of the call integration system 106, which determines output genotype calls based on genotype probabilities or variant call classifications from a genotype call integration machine learning model. Additional details regarding the operation of FIG. 2 are provided subsequently with reference to subsequent figures.

[0061] As shown in FIG. 2 , the call integration system 106 performs operation 202 to receive a first genotype call and a second genotype call. In particular, in some embodiments, the call integration system 106 receives the first genotype call represented by a first VCF file generated from nucleotide reads of a first read type. Additionally, the call integration system receives the second genotype call represented by a second VCF file generated from nucleotide reads of a second read type. In some cases, the call integration system 106 generates the first genotype call by analyzing SBS reads, e.g., nucleotide reads synthesized from sample library fragments shorter than a first threshold number of nucleobases. In these or other cases, the call integration system 106 generates the second genotype call by analyzing different types of read data, e.g., (i) assembled nucleotide reads—nucleotide reads assembled from shorter nucleotide reads to form contiguous sequences, (ii) CCS reads, and / or (iii) nanopore long reads. In certain embodiments, the first received genotype call and the second received genotype call are initial genotype calls that the call integration system 106 uses as a basis to ultimately generate the output genotype call (e.g., by merging data associated with the first genotype call and the second genotype call).

[0062] 2 , the call integration system 106 performs operation 204 to identify sequencing metrics. In particular, the call integration system 106 identifies or determines sequencing metrics, such as read-based sequencing metrics, external source sequencing metrics, and call model generation sequencing metrics. For example, the call integration system 106 determines sequencing metrics that indicate various attributes or data regarding various genotype calls of nucleotide reads from a sample nucleotide sequence. In some embodiments, the call integration system 106 determines or extracts different sequencing metrics for generating genotype calls associated with different variants, such as SNPs and indels. Indeed, from the different sequencing metrics, the call integration system 106 can generate output genotype calls corresponding to each variant on which the genotype call integration machine learning model is trained.

[0063] 2 , the call integration system 106 utilizes different instances of a genotype call integration machine learning model to generate different predictions for different variants based on the extracted sequencing metrics. For example, to generate an output genotype call corresponding to a (biallelic) SNP, the call integration system 106 performs operation 206 to generate a genotype probability. As another example, to generate an output genotype call corresponding to an indel (or a multi-allelic SNP, or a variant other than a biallelic SNP), the call integration system 106 performs operation 208 to generate a variant call classification. As described below, in some embodiments, the call integration system 106 may use one or both of a SNP-specific genotype call integration machine learning model to generate genotype probabilities and an indel-specific genotype call integration machine learning model to generate a variant call classification. In some cases, the call integration system 106 may analyze or determine biallelic SNPs using a biallelic SNP genotype call integration machine learning model. In these or other cases, the call integration system 106 may use an indel-specific genotype call integration machine learning model to analyze or determine indels, multi-allelic SNPs, or other variants that are not bi-allelic SNPs.

[0064] To generate genotype probabilities (e.g., via operation 206), the call integration system 106 utilizes a genotype call integration machine learning model to analyze sequencing metrics (e.g., SNP-related sequencing metrics). Specifically, the call integration system 106 utilizes a genotype call integration machine learning model trained on the SNP training data to generate genotype probabilities for one or more candidate SNPs. From the sequencing metrics, the genotype call integration machine learning model generates a set of genotype probabilities for a particular genomic coordinate, indicating the likelihood of a 0 / 0 genotype (e.g., a homozygous reference genotype), the likelihood of a 0 / 1 genotype or a 1 / 0 genotype (e.g., a heterozygous genotype), and the likelihood of a 1 / 1 genotype (e.g., a homozygous alternative genotype).

[0065] To generate variant call classifications (e.g., via operation 208), the call integration system 106 utilizes a genotype call integration machine learning model to generate (or update or refine) variant call classifications from sequencing metrics. Specifically, the call integration system 106 utilizes the genotype call integration machine learning model to process or analyze one or more sequencing metrics and generate a set of classifications (e.g., predicted probabilities associated with variant, zygosity, or reference calls). For example, the call integration system 106 utilizes the genotype call integration machine learning model to generate a set of variant call classifications including: i) a first true-positive variant probability for the first genotype call (e.g., from the first read type), ii) a second true-positive variant probability for the second genotype call (e.g., from the second read type), iii) a first zygosity error probability for the first genotype call, iv) a second zygosity error probability for the second genotype call, and v) a reference probability.

[0066] As further illustrated in FIG. 2 , the call integration system 106 also performs operation 210 of generating an output genotype call. In particular, the call integration system 106 generates an output genotype call for one or more genomic coordinates of a SNP based on the genotype probability output by the genotype call integration machine learning model. Additionally, or alternatively, the call integration system 106 generates an output genotype call for one or more genomic coordinates of an indel based on the variant call classification output by the genotype call integration machine learning model. For either a SNP or an indel, the call integration system 106 determines or updates the genotype call by merging, or otherwise generating a merged VCF file from, data associated with a first read type (e.g., an SBS read) and data associated with a second read type (e.g., an assembled nucleotide read). In some cases, the call integration system 106 determines an output genotype indicating the presence or absence of a SNP or indel at one or more genomic coordinates of the genomic sample. For example, the call integration system 106 selects the initial genotype call (e.g., the first genotype call or the second genotype call) as the output genotype call. Alternatively, the call integration system 106 generates an output genotype call that is different from the initial genotype call (e.g., the first genotype call and the second genotype call) but is based on data associated with the initial genotype call.

[0067] In some embodiments, the call integration system 106 utilizes a call generation model to generate a merged VCF file from the genotype probabilities and / or variant call classifications (as generated by the genotype call integration machine learning model). For example, the call integration system 106 applies a number of Bayesian probability models or algorithms to derive various probabilities for different nucleobases, quality metrics, mapping metrics, joint metrics, and other data occurring in the sample nucleotide sequence for inclusion in the variant call file. From the probability models, the call integration system 106 can further determine output genotype calls indicating predicted genotypes (or variants) of the sample genome at genomic coordinates corresponding to the reference genome.

[0068] As part of generating the output genotype call, in certain implementations, the call integration system 106 utilizes the genotype probabilities and / or variant call classifications to generate, recalibrate, determine, correct, confirm, or augment the initial genotype call(s). Specifically, the call integration system 106 utilizes the genotype probabilities and / or variant call classifications (and / or other features) to determine or update certain metrics associated with the genotype calls. For example, the call integration system 106 modifies data fields corresponding to the variant call file for metrics such as call quality, genotype, and genotype quality (or others, as described below) to generate the output genotype call (e.g., as a new genotype call or as a modified or merged version of the first and / or second genotype calls).

[0069] While FIG. 2 illustrates a particular order for operations 202-210, in some embodiments, the call integration system 106 performs the operations simultaneously, in a different order, and / or in parallel. For example, the call integration system 106 may perform operation 206 for generating genotype probabilities and / or operation 208 for generating variant call classifications while or during the process of performing operation 210 for generating output genotype calls. For example, the call integration system 106 may simultaneously implement a genotype call integration machine learning model and a call generation model to generate output genotype calls and genotype probabilities / variant call classifications for modifying the output genotype calls. In some cases, the call integration system 106 further modifies data fields corresponding to the merged variant call file (e.g., in a pre-filter or post-filter variant call file) of the output genotype calls. As alluded to above, this simultaneous or parallel operation provides the call integration system 106 with improved computational efficiency and increased speed by recalibrating genotype calls when they are initially generated (rather than performing one operation before the other).

[0070] In one or more implementations, the call consolidation system 106 determines an output genotype call as part of a genomic coordinate(s) tagged for the SNP or indel. For example, the call consolidation system 106 determines an output genotype call representing a SNP at a genomic coordinate (e.g., chr1:151863125) by identifying a G in the sample nucleotide sequence where an A is present in the reference genome. As another example, the call consolidation system 106 determines that a genotype call around one or more genomic coordinates (e.g., chr1:49263256) represents a deletion by identifying a single G in the sample nucleotide sequence where a GTAAC is present in the reference genome. As a further example, the call consolidation system 106 determines that a sequence of a genotype call represents an insertion at a genomic coordinate (e.g., chr1:7602080) by identifying a sequence of TTTCC in the sample nucleotide sequence where a T is present in the reference genome. Indeed, in some cases, an insertion comprises a sequence of genotype calls that replaces a single reference base in the genomic coordinates of the reference sequence.

[0071] As noted above, in certain embodiments, the call integration system 106 receives, identifies, or determines initial genotype calls from different types of nucleotide reads. In particular, the call integration system 106 utilizes a multi-read pipeline to merge sequencing metrics or other data from one type of nucleotide read (e.g., short reads or SBS reads) with sequencing metrics or other data from another type of nucleotide read (e.g., long reads or assembled nucleotide reads) to generate an output genotype call from the initial genotype call. Figure 3 illustrates exemplary types of nucleotide reads for which the call integration system 106 may analyze or receive associated data as part of generating an output genotype call at genomic coordinates, according to one or more embodiments. As indicated above, in some embodiments, the call integration system 106 identifies or determines genotype calls and corresponding sequencing metrics based on the first type of nucleotide read and the second type of nucleotide read shown in Figure 3.

[0072] As illustrated in FIG. 3 , the call integration system 106 analyzes read data associated with a first type of nucleotide read 302. In particular, the call integration system 106 receives or determines a first genotype call from the first type of nucleotide read 302. For example, the call integration system 106 determines or receives a genotype call indicative of a genotype or mutation at a particular genomic coordinate as indicated by the read of the first type of nucleotide read 302. In some embodiments, the first type of nucleotide read 302 includes short reads (e.g., reads shorter than a threshold length or composed of fewer than a threshold number of nucleobases), such as SBS reads synthesized from sample library fragments shorter than a threshold number of nucleobases. In certain embodiments, the call integration system 106 determines the first type of nucleotide read 302 from wells in a flow cell and / or via fluorescent tagging. In some cases, the call integration system 106 utilizes cluster generation and SBS chemistry to sequence millions or billions of clusters in a flow cell. During SBS chemistry, for each cluster, the call integration system 106 stores nucleobase calls from the nucleotide reads for each sequencing cycle via Real-Time Analysis (RTA) software. While the particular genomic coordinates above include a first genotype call based on a read of a first type of nucleotide read 302 and a second genotype call based on a read of a second type of nucleotide read 304, the particular genomic coordinates may include only a genotype call based on a read of a first type of nucleotide read 302 or only a genotype call based on a read of a second type of nucleotide read 304, but not both.

[0073] 3 , the call integration system 106 analyzes read data associated with the second type of nucleotide reads 304. In particular, the call integration system 106 receives or determines a second genotype call from the second type of nucleotide reads 304. For example, the call integration system 106 determines or receives a genotype call indicative of a genotype or mutation at a particular genomic coordinate as indicated by the read of the second type of nucleotide reads 304. More specifically, the second type of nucleotide reads 304 may include assembled nucleotide reads, CCS reads, and / or long reads (e.g., reads longer than a threshold length or composed of at least a threshold number of nucleobases), such as nanopore long reads.

[0074] With respect to assembled nucleotide reads, the call integration system 106 may determine assembled nucleotide reads by utilizing a mutagenesis process and a rendering process. Specifically, the call integration system 106 may assemble, create, synthesize, or generate long reads from short reads. For example, the call integration system 106 may apply mutations to a set of short reads (e.g., SBS reads or other short reads) to introduce unique genetic signatures, thereby enabling assembly to work across low-complexity regions with many repeats. In some cases, the call integration system 106 applies random mutations and uses the output of the mutated short reads to recover information in regions of the sample genome that are difficult to sequence using conventional SBS techniques. For example, the call integration system 106 may combine the mutated short reads to form assembled long reads, and the call integration system 106 may further perform a rendering process to reverse at least some of the mutations after the short reads have been combined or assembled into long reads.

[0075] As noted above, in certain described embodiments, the call integration system 106 determines or extracts sequencing metrics for genotype calls at genomic coordinates. In particular, the call integration system 106 determines sequencing metrics, such as read-based sequencing metrics, external source sequencing metrics, and call model generation sequencing metrics, from calls corresponding to nucleotide reads from a sample nucleotide sequence. Figures 4A-4C illustrate the call integration system 106 determining sequencing metrics, according to one or more embodiments. Specifically, Figure 4A illustrates the call integration system 106 determining read-based sequencing metrics based on a first type of nucleotide read and a second type of nucleotide read. FIG. 4B illustrates the call integration system 106 determining sequencing metrics for call model generation for genotype calls corresponding to either a first type of nucleotide read or a second type of nucleotide read, and FIG. 4C illustrates the call integration system 106 identifying or determining external source sequencing metrics for genomic coordinate(s) of genotype calls corresponding to either a first type of nucleotide read or a second type of nucleotide read.

[0076] 4A , the call consolidation system 106 accesses, retrieves, acquires, determines, receives, or generates nucleotide reads, including a first type of nucleotide read 402a (e.g., first type of nucleotide read 302) and a second type of nucleotide read 402b (e.g., second type of nucleotide read 304). For example, the call consolidation system 106 utilizes a sequencing device 114 to determine nucleotide reads for regions from a sample nucleotide sequence (e.g., a sample genome). For example, the call consolidation system 106 utilizes sequencing-by-synthesis (SBS) technology, Sanger sequencing technology, assembled nucleotide read technology, or other sequencing technology discussed herein to generate multiple nucleotide reads and determine genotype calls for oligonucleotide clusters.

[0077] As further illustrated in FIG. 4A , in some embodiments, the call integration system 106 performs read processing and mapping 404a on the first type of nucleotide reads 402a and performs read processing and mapping 404b on the second type of nucleotide reads 402b. For example, the call integration system 106 utilizes RTA software to store the base call data in the form of individual base call data files (or BCLs). In some cases, the call integration system 106 further converts the BCL files (e.g., via a BCL-to-FASTQ conversion) into sequence data 408a and 408b, as illustrated in FIG. 4B , where sequence data 408a corresponds to the first type of nucleotide reads 402a and sequence data 408b corresponds to the second type of nucleotide reads 402b.

[0078] As illustrated in FIG. 4A , the call consolidation system 106 generates multiple read coverage (e.g., a read pileup) that includes multiple nucleotide reads or nucleobase calls corresponding to a single genomic coordinate. In particular, in certain embodiments, the call consolidation system 106 aligns nucleotide reads to a reference genome or receives information regarding read alignment. Specifically, the call consolidation system 106 determines (or receives information indicating) which nucleobase(s) of a given read align with which genomic coordinate of a reference sequence. Different reads have different lengths and contain different nucleobases. Thus, in some cases, the call consolidation system 106 analyzes each nucleotide of each read to determine (or receives information indicating) where the read "fits" with respect to the reference sequence, e.g., where bases in the read align with bases in the reference. In some cases, the call consolidation system 106 aligns many reads at a single genomic coordinate, thus resulting in a read pileup.

[0079] In certain embodiments, the call integration system 106 performs additional statistical tests to determine or detect differences between the metrics associated with the reference nucleotide sequence and the metrics associated with the alternative supporting nucleotide reads. Through these statistical tests, the call integration system 106 rescales the raw sequencing metrics to determine read-based sequencing metrics 406a for the first type of nucleotide read 402a and read-based sequencing metrics 406b for the second type of nucleotide read 402b. In some embodiments, the call integration system 106 determines a shared set of sequencing metrics associated with both the first type of nucleotide read 402a and the second type of nucleotide read 402b.

[0080] In some cases, the call consolidation system 106 determines or extracts raw sequencing metrics, including one or more of: (i) an alignment metric to quantify the alignment of a sample nucleotide sequence with the genomic coordinates of an exemplary nucleotide sequence (e.g., a nucleotide sequence from a reference genome or an ancestral haplotype); (ii) a depth metric to quantify the depth of a nucleobase call for the sample nucleotide sequence at the genomic coordinates of the exemplary nucleotide sequence; or (iii) a call quality metric to quantify the quality of a nucleobase call for the sample nucleotide sequence at the genomic coordinates of the exemplary nucleotide sequence. For example, the call consolidation system 106 determines a mapping quality metric (e.g., a MAPQ metric), a soft clipping metric, or other alignment metric that measures the alignment of the sample sequence with the reference genome. As another example, the call consolidation system 106 determines a forward or reverse depth metric (or other such depth metric) or a callability metric (or other such call quality metric) for a genotype call or variant call.

[0081] As just mentioned, in some embodiments, the call integration system 106 re-manipulates raw sequencing metrics to generate read-based sequencing metrics 406a and 406b that are more useful for comparing metrics associated with a reference nucleotide sequence with metrics associated with various supporting alternative nucleotide reads. For example, the call integration system 106 determines various metrics for a sample sequence relative to a reference sequence and also determines various metrics for a sample sequence relative to an alternative supporting sequence. Note that the call integration system 106 performs a comparative analysis between the metrics associated with the reference sequence and the metrics associated with the alternative supporting reads.

[0082] For example, the call integration system 106 compares the mapping of nucleobases of a sample nucleotide sequence (e.g., a sample genome) to a reference sequence with the mapping of nucleobases to various alternative supporting reads. In some cases, the call integration system 106 determines a mapping quality associated with the reference sequence to compare with the mapping quality associated with the alternative supporting reads. For example, the call integration system 106 determines a mapping quality statistic that reflects the difference in the distribution of reads supporting the reference sequence versus reads supporting alternative alleles.

[0083] In these or other cases, the call integration system 106 determines mismatch counts between the sample sequence and the reference sequence, and between the reference sequence and the alternate support read. The call integration system 106 further compares the mismatch counts to determine a comparative mismatch count metric. The call integration system 106 further determines a soft clipping metric for the sample sequence relative to the reference sequence, and further determines a soft clipping metric for the alternate support read. The call integration system 106 also compares the soft clipping metrics between the reference sequence and the alternate support read to generate a comparative soft clipping metric. Furthermore, the call integration system 106 compares base call quality metrics for the reference sequence and the alternate support read, and / or compares the query position of the sample sequence relative to the reference sequence with the query position for the alternate support read.

[0084] 4A, the call integration system 106 utilizes comparisons and / or other statistical tests to generate the read-based sequencing metrics 406a and 406b. In some cases, the call integration system 106 generates the read-based sequencing metrics 406a and 406b to include one or more of the same metrics listed above. For example, from the first type of nucleotide reads 402a and the second type of nucleotide reads 402b, the call integration system 106 may generate: i) an allele frequency metric indicating the frequency of occurrence of the alleles of the first genotype call, the alleles of the second genotype call, or the different alleles of an alternative genotype call that is different from the first genotype call and the second genotype call; ii) a coverage depth metric indicating a particular (e.g., maximum or cumulative total) coverage depth for the first type of nucleotide reads 402a corresponding to the first genotype call or the second type of nucleotide reads 402b corresponding to the second genotype call; and iii) a coverage depth metric indicating the coverage depth of the first type of nucleotide reads 402a corresponding to the first genotype call or the second type of nucleotide reads 402b corresponding to the second genotype call. iv) a mapping quality metric (e.g., a MAPQ score) for the second type of nucleotide reads 402b corresponding to the first or second genotype call; iv) a nucleobase composition metric indicating the nucleobase configuration or composition at the genomic coordinates of one or more nucleotide reads from the first type of nucleotide reads 402a or the second type of nucleotide reads 402b; and v) a mean coverage depth metric indicating the average (e.g., mean or median) coverage depth for the first type of nucleotide reads 402a corresponding to the first genotype call or the second type of nucleotide reads 402b corresponding to the second genotype call.

[0085] The call integration system 106 also generates read-based sequencing metrics 406b from the second type of nucleotide reads 402b, which may not be applicable to the first type of nucleotide reads 402a, using comparisons and statistical tests. For example, the call integration system 106 generates read-based sequencing metrics 406b including: i) an assembly score, which indicates a measure of accuracy or completeness for the assembled reads generated using mutagenesis and rendering; ii) k-mer statistics, which indicate the length of the read and / or the length of the mutations (e.g., insertions or deletions); and iii) rendering metrics, which indicate a measure of accuracy or completeness of the rendering of mutations (e.g., from the mutagenesis process) from the assembled nucleotide reads. Further details regarding the read-based sequencing metrics 406a and 406b are provided below.

[0086] A. Read-Based Sequencing Metrics The following paragraphs describe various read-based sequencing metrics in more detail. For example, the call integration system 106 determines a base call quality score for a base call within a nucleotide read. Specifically, the call integration system 106 determines the probability of accuracy of a nucleobase call of a nucleotide read (e.g., encoded PHRED+33). In some cases, the call integration system 106 determines one or more base call quality scores in the form of a DRAGEN QUAL score or Q score for one or more nucleobase calls. Additionally, the call integration system 106 determines the percentage of nucleotide reads that support alternative contiguous sequences from the reference genome. For example, the call integration system 106 determines the number of nucleotide reads that support (e.g., match or align with) alternative contiguous sequences of the reference genome and the number of nucleotide reads that support the primary assembly within the reference genome. The call integration system 106 further compares the aforementioned numbers and determines a percentage to reflect the comparison.

[0087] In some cases, the call integration system 106 determines the percentage of reads that support alternative contiguous sequences using certain features including: i) alignment scores with respect to the reference genome; ii) alignment scores with respect to the assembly of alternative contiguous sequences; iii) mapping quality of the nucleotide reads; and iv) the amount of overlap with the genomic region. Furthermore, the call integration system 106 may classify reads based on their alignment according to the following categories: i) perfect alignment to the alternative contiguous sequence assembly (e.g., meeting a first alignment score threshold), ii) perfect alignment to the reference genome, iii) strong alignment to the alternative contiguous sequence assembly (e.g., meeting a second alignment score threshold but not meeting the first alignment score threshold), iv) strong alignment to the reference genome (e.g., also meeting a second alignment score threshold but not meeting the first alignment score threshold), and v) no strong alignment to either the alternative contiguous sequence assembly or the reference genome (e.g., not meeting the second alignment threshold for both the alternative contiguous sequence assembly and the reference genome). Based on these five categories, the call integration system 106 may further determine a proportion comparing each of these categories to determine the proportion of nucleotide reads that support the alternative contiguous sequence (e.g., the proportion of reads that overlap with the target genomic region) versus the proportion of nucleotide reads that support the reference genome.

[0088] Note that the call integration system 106 may determine the number of split nucleotide reads from nucleotide reads corresponding to a genotype call or a variant call as a read-based sequencing metric. More specifically, the call integration system 106 determines the number of nucleotide reads that do not have a continuous alignment with the primary assembly of the reference genome (or that align less than a threshold number of bases), but rather contain nucleotide read fragments that align with two or more reference sequences in the reference genome. For example, the call integration system 106 uses a call generation model to determine split read counts that support a genotype call. For heterozygous deletion calls, some false-positive cases have a higher-than-expected coverage depth and a large split read count that exceeds the counts in true-positive cases. Therefore, the call integration system 106 may generate a split nucleotide read metric based on the nucleotide reads that support the genotype call.

[0089] In some embodiments, the call integration system 106 compares split read evidence supporting alternative alleles for each forward and reverse nucleotide read. If most of the evidence is from either the forward or reverse read, this bias may indicate a systematic problem, especially when the read count is relatively high (e.g., greater than 10 nucleotide reads). The call integration system 106 uses the forward and reverse read counts with perfect alignment scores to the contiguous sequence as the sequencing metric for the genotype call integration machine learning model.

[0090] As described above, the call integration system 106 may determine the coverage depth of nucleotide reads corresponding to the initial structural variant calls as a read-based sequencing metric. For example, the call integration system 106 may determine the count or number of nucleotide reads that overlap with the target genomic region corresponding to a variant identified as present or absent by the initial genotype or variant call. Thus, the coverage depth may be represented by the raw count of nucleotide reads that overlap with the target genomic region by at least a threshold number of nucleotide bases.

[0091] Additionally, the call integration system 106 may determine additional genotype calls (e.g., variant calls) within a threshold number of base pairs from an initial genotype call (e.g., variant call) within the genomic sample as part of the read-based sequencing metrics. For example, the call integration system 106 may determine variant calls such as insertions or deletions within a threshold proximity (e.g., within 200 base pairs) of the initial variant call. Thus, the call integration system 106 may indicate the presence or absence of such additional variant calls using a code, such as a binary code of 0 representing absence and 1 representing presence.

[0092] In some embodiments, the call integration system 106 further determines, as a read-based sequencing metric, an alignment of a contiguous sequence corresponding to the nucleotide read with a reference sequence of a reference genome modified to include variations corresponding to the initial genotype call. In particular, the call integration system 106 modifies the reference genome by changing nucleotide bases to reflect variations, such as SNPs and indels, in adjacent regions. In theory, the modified reference genome could perfectly align with the alternate contiguous sequence, providing some training advantage to the genotype call integration machine learning model in accurately identifying variations.

[0093] To modify the reference genome to include a mutation, the call integration system 106 may perform various steps. In particular, the call integration system 106 may remove a portion of the sequence from the reference genome that corresponds to a deleted region (e.g., the deleted region of a deletion mutation). In some cases, the call integration system 106 may replace the relevant portion of the reference sequence in a FAST-All (FASTA) file with a contiguous sequence representing the relevant mutation. The call integration system 106 may then regenerate the hash table using the modified FASTA file. Additionally, the call integration system 106 may run the mapping and alignment component of the call generation model on the modified reference genome. The call integration system 106 may further rerun the mutation caller component of the call generation model on the new mapping and alignment output.

[0094] For candidate mutations for which read-based evidence falls below a threshold (e.g., fewer than 5 or 10 nucleotide reads supporting the candidate mutation call), one approach to finding missing reads is to modify the local reference sequence by replacing it with a contiguous sequence representing the candidate mutation. In the case of a true positive, when the reads are remapped to the modified reference genome, some of the nucleotide reads that were incorrectly mapped / aligned to the primary assembly of the reference genome are more likely to map correctly to the contiguous sequence representing the candidate mutation, thereby increasing the read depth on the new modified reference genome. Based on the new mapping, when the call integration system 106 reruns the call generation model, the call generation model will not call a mutation for a true homozygous deletion or an insertion for a true heterozygous deletion. Furthermore, the read coverage depth should increase for the contiguous sequence representing the candidate mutation compared to the original primary assembly, which should result in more accurate mutation calls. The likelihood of achieving more accurate mapping can be estimated by aligning read-length segments of the contiguous sequence representing the candidate mutation to the reference genome.

[0095] In some embodiments, the call integration system 106 analyzes flanking regions of the variant (as called by the call generation model) in the sample sequence, where the flanking regions include base calls within a threshold proximity (e.g., within 200 base pairs) of the variant. For example, the call integration system 106 uses a call generation model (e.g., DRAGEN VC) to determine an initial variant based on an initial genotype call, modifies the reference genome to include (a portion of) a contiguous sequence reflecting the variant, and identifies flanking regions of a threshold size of 200 base pairs on either side of the variant. The call integration system 106 further analyzes flanking regions (e.g., left and right flanking regions) of the combined sequence to determine the presence or absence of the variant. Indeed, the call integration system 106 may quantify the extent (e.g., amount, magnitude, and / or size) of single nucleotide polymorphisms (SNPs) and / or insertions or deletions (indels) based on the modified reference genome (e.g., the combined sequence of the reference genome and the contiguous sequence).

[0096] In some cases, interpretation of contiguous sequences is sensitive to the scoring parameters and penalties within the Smith-Waterman algorithm. Thus, in these or other cases, the call consolidation system 106 may use deletion counts from the Concise Idiosyncratic Gapped Alignment Report (CIGAR) string output for multiple scoring parameter sets to measure sensitivity to Smith-Waterman scoring parameters / penalties. The call consolidation system 106 may further use the maximum contiguous deletion length, as well as the sum of all deletions corresponding to genomic regions spanned by interchromosomal translocations, as a sequencing metric (e.g., a read-based sequencing metric).

[0097] In some cases, the call integration system 106 determines read-based sequencing metrics in the form of deletion lengths in nucleotide bases based on one or more soft-clipped nucleotide reads. For example, the call integration system 106 realigns soft-clipped segments from nucleotide reads to determine deletion lengths (or lengths of different types of mutations). In some embodiments, the call integration system 106 realigns only the soft-clipped portion of the read to provide an estimate of the length of the deletion or some other mutation. For example, the call integration system 106 performs realignment only if the size of the soft-clipped portion meets (e.g., is greater than) a threshold number of soft-clipped bases (e.g., 10 soft-clipped bases or 20 soft-clipped bases).

[0098] Note that in some embodiments, the call integration system 106 determines or calculates realignment offsets for soft-clipped segments (e.g., segments that meet the length requirement) by: i) for reads that are soft-clipped to the left of the called mutation, aligning the soft-clipped portion to the left of the current position / coordinate indicating the end of the soft clip; ii) for reads that are soft-clipped to the right of the current position / coordinate indicating the start of the soft clip; iii) determining the distance in number of nucleotide bases between the aligned position / coordinate and the position of the soft clip from the original mapping; iv) determining left and right modes for all distances determined through steps i) to iii); and v) determining or calculating left realignment offsets and right realignment offsets by determining the difference between the left mode and the deletion length determined by the call generation model (e.g., DRAGEN SV Caller) and between the right mode and the deletion length determined by the call generation model (e.g., DRAGEN SV Caller), e.g., the number of nucleotide bases determined from the mutation length - alt seq length.

[0099] Additionally, the call integration system 106 may determine read-based sequencing metrics in the form of the number of nucleotide reads exhibiting a mapping quality metric that does not meet a threshold mapping quality metric. Specifically, the call integration system 106 corrects for cases where a true positive indicates a nucleotide read with a low MAPQ score (i.e., below a threshold MAPQ) but is still correctly mapped (although possibly locally incorrect). In some cases, the call integration system 106 utilizes MAPQ as a soft weighting to indicate the likelihood of aligning with an alternative contiguous sequence or reference genome. The call integration system 106 may further determine a count or number of reads with a mapping quality metric (e.g., MAPQ score) that does not meet (or is below) a threshold mapping quality metric (e.g., MAPQ=10 or MAPQ=60 or a relative MAPQ threshold). In some cases, the call integration system 106 determines or generates a variant call based on the number of reads with a low mapping quality metric. In certain embodiments, such as when MAPQ=60, the call integration system 106 further incorporates an XQ score to determine an extended range for the likelihood of the variant. The call integration system 106 may determine and incorporate the standard deviation of XQ across locally mapped reads to improve the predictions of the genotype call integration machine learning model.

[0100] As further described above, in some embodiments, the call integration system 106 also determines an insert size, which represents the length of a nucleotide read fragment corresponding to the initial genotype or variant call determined by the call generation model. Specifically, the call integration system 106 determines the size or length (e.g., number of base pairs) of an insertion (or other variant) within a genomic region (e.g., an SV region) of the genomic sample.

[0101] In some cases, the call integration system 106 determines read-based sequencing metrics in the form of palindrome (self-complementary sequence, palindrome) metrics. For example, the call integration system 106 analyzes portions of reference sequences corresponding to target genomic regions where variants are called (e.g., by a call generation model). Specifically, if a reference sequence in such a target genomic region is palindromic (or within a threshold percentage of palindromes, or within a threshold number of base pairs from a palindrome), the likelihood of a folding effect increases. Based on the analysis, the call integration system 106 identifies or detects fragments or portions of the genomic sample (e.g., subsequences of reads) that are within a threshold distance (e.g., within 200 base pairs) from each other and are palindromic (which may indicate deletions due to a folding effect during base calling). The call integration system 106 may determine or measure the distance or proximity (e.g., the number of base pairs separating) of segments of the palindrome metric. In some cases, the call integration system 106 further incorporates permutation entropy into the palindrome metric, such that palindromic matches (e.g., pairs of segments that are palindromic to each other) with higher permutation entropy increase the likelihood of a deletion (or some other mutation).

[0102] Furthermore, in some embodiments, the call integration system 106 determines a read-based sequencing metric in the form of a mutation likelihood or probability, which represents the ratio of an initial mutation call to a reference call for one or more genomic coordinates based on insert size. In particular, assuming no mutation is present, there is a certain implied insert size or fragment size. On the other hand, assuming a mutation is present, there is a different implied insert size or fragment size. Thus, based on the mean and standard deviation of the fragment sizes, the call integration system 106 may determine whether the presence or absence of a mutation is more likely. For example, in some embodiments, the call integration system 106 determines the ratio of an initial mutation call to a reference call for one or more genomic coordinates according to the following formula:

[0103]

number

[0104]

number

[0105]

number

[0106] Depending on the read orientation and alignment to the candidate variant genomic region, the call integration system 106 may subtract the length of the proposed variant (e.g., deletion) from the original insert size estimate (e.g., based on reference mapping and alignment). When considering all nucleotide reads that provide evidence supporting the alternative allele, the call integration system 106 may determine a likelihood ratio (e.g., alt vs. ref) based on the predicted insert size across the set of reads.

[0107] In some cases,

[0108]

number

[0109] In one or more embodiments, the call integration system 106 determines read-based sequencing metrics in the form of confidence intervals around the end breakpoints. In particular, the call integration system 106 utilizes a call generation model to determine the confidence interval as a measure of the certainty of the breakpoint location. For example, the call integration system 106 determines a range of reference coordinates within which the breakpoint may be located corresponding to the variant call. In some cases, the call integration system 106 determines the range of reference coordinates to reflect a threshold percentile (e.g., the 95th percentile) for the confidence interval.

[0110] In certain embodiments, the call integration system 106 further determines additional or alternative read-based sequencing metrics. For example, the call integration system 106 determines the length of homology as a read-based sequencing metric. Specifically, the call integration system 106 determines the length of nucleotide base sequences repeated in the target genomic region of the variation and / or the length of nucleotide base sequences having at least a threshold measure of homology with other nucleotide base sequences (of similar length) within the target genomic region of the structural variation (e.g., HOMLEN=8 GCTTGAAC GCTTAAAC GCTAGAAC GCTTGAAC GCTTGTAC, etc.). In some cases, the call integration system 106 determines the length of an inserted nucleotide base sequence as a read-based sequencing metric. In these or other cases, the call integration system 106 determines the homology of the inserted nucleotide base sequence to a reference sequence within the target genomic region of the variation.

[0111] In one or more embodiments, the call integration system 106 determines additional or alternative read-based sequencing metrics, including: i) a comparative mapping quality distribution metric indicating a mapping quality distribution comparing mapping quality for the reference sequence to mapping quality for the alternative supporting read; ii) a comparative secondary mapping alignment metric indicating a comparison of secondary mapping for bases in the reference sequence to secondary mapping for bases in the alternative supporting read; iii) a comparative mismatch count metric indicating a comparison of mismatched nucleobases for the reference sequence to mismatched bases for the alternative supporting read; iv) a comparative soft clipping metric indicating a comparison of a soft clipping metric for the reference sequence to a soft clipping metric for the alternative supporting read; v) a read depth of the nucleotide read and one or more averages vi) one or more comparative base quality metrics indicating a comparison between base quality for the reference sequence and base quality for the alternative supporting reads (e.g., overall base quality, initial base quality, and final base quality in the nucleotide read); vii) comparative query position metrics indicating a comparison between a query position for the reference sequence and a query position for the alternative supporting reads; viii) one or more context information metrics indicating homopolymer and periodicity of the nucleobase calls; ix) a strand bias metric indicating a strand bias associated with one or more of the nucleotide reads; and x) a read direction bias metric indicating a read direction bias associated with the nucleotide reads.

[0112] B. Sequencing metrics for call model generation As shown in FIG. 4B , in addition to read-based sequencing metrics 406a and 406b, the call integration system 106 generates call-model-generated sequencing metrics 412a and 412b. In particular, the call integration system 106 generates call-model-generated sequencing metrics 412a and 412b from sequence data 408a and 408b, respectively, utilizing instances of call generation models 410a and 410b. For example, the call integration system 106 extracts or determines sequence data 408a based on read processing and mapping 404a, as described in connection with FIG. 4A . Similarly, the call integration system 106 extracts or determines sequence data 408b based on read processing and mapping 404b. In some cases, the call integration system 106 generates sequence data 408a and 408b as part of one or more digital files, such as BCL and FASTQ files.

[0113] To generate such files, in some embodiments, the sequencing device 114 (or call integration system 106) sequences millions or billions of clusters in a flow cell using cluster generation and SBS chemistry. During SBS chemistry, for each cluster, the sequencing device 114 (or call integration system 106) stores nucleobase calls from the first type of nucleotide reads 402a and the second type of nucleotide reads 402b for each sequencing cycle via real-time analysis (RTA) software. The sequencing device 114 (or call integration system 106) further stores the base call data in the form of individual base call data files (or BCLs) using the RTA software. In some cases, the sequencing device 114 (or call integration system 106) further converts the BCL files (e.g., via BCL-to-FASTQ conversion) into sequence data 408a and 408b. For example, the sequencing device 114 (or call integration system 106) generates FASTQ files from a first type of nucleotide read 402a and a second type of nucleotide read 402b, which contain sequence data 408a and 408b, respectively.

[0114] In some cases, the call integration system 106 generates sequence data 408a and 408b for each cluster that passes the initial quality filter from the sample sequence. For example, the call integration system 106 generates an entry for each cluster, with each entry including four rows (or four items of sequence data): i) a sequence identifier with information about the sequencing run and the cluster, ii) the nucleobase calls (e.g., a sequence of A, C, T, G, and / or N calls) that make up the sequence, iii) a separator (e.g., a "+" symbol), and iv) a base call quality metric indicating the probability of accuracy for the nucleobase call (encoded PHRED+33).

[0115] 4B , the call integration system 106 implements, utilizes, or applies a call generation model 410a to process or analyze the sequence data 408a. Similarly, the call integration system 106 implements, utilizes, or applies a call generation model 410b to process or analyze the sequence data 408b. Indeed, in some embodiments, the call integration system 106 utilizes respective instances of the call generation models 410a and 410b to generate call-model-generated sequencing metrics 412a and 412b by re-manipulating raw sequencing metrics (e.g., the raw sequencing metrics in the sequence data 408a and 408b). In particular, the instances of the call generation models 410a and 410b include mapping and alignment components for mapping and aligning nucleobase calls from the sequence data 408a and 408b. Note that the instances of the call generation models 410a and 410b include a variant calling component for generating an initial genotype call (e.g., a reference base call such as a nucleobase call, a variant call, or a non-variant call) from the sequence data 408a and 408b. In some cases, the call integration system 106 extracts sequencing metrics 412a and 412b for generating the call models, which have been generated using the mapping and alignment component and the variant calling component of the instances of the call generation models 410a and 410b.

[0116] To illustrate examples of call model generation sequencing metrics 412a and 412b, in some cases, the call integration system 106 may generate: i) a genotype metric that corresponds to a GT field in a VCF file and indicates the genotype of the genomic coordinate; ii) a base call quality metric (e.g., a DRAGEN QUAL score) that indicates a quality score of the genotype call generated via the call generation model 410a or 410b; iii) a genotype quality metric (e.g., a GQ score) that indicates a measure of the reliability or quality of the genotype predicted for the genomic coordinate; iv) a genotype probability metric that indicates one or more probabilities that various genotypes will occur at the genomic coordinate; v) a PHRED-scaled likelihood metric or a non-PHRED-scaled likelihood metric that indicates the probability of an error associated with a genotype call; and vi) a call model generation foreign read detection metric (e.g., a Foreign Read Detection metric) that indicates the probability that one or more of the first type nucleotide reads 402a or the second type nucleotide reads 402b in the pileup are foreign reads (e.g., their true location is elsewhere in the reference sequence). vii) a call model-generated base quality drop metric (e.g., a Base Quality Dropoff (FRD) score) indicating a probability of base quality drop based on one or more of strand bias, error location within a thread, or low average base quality across a subset of the first type nucleotide reads 402a and / or the second type nucleotide reads 402b; viii) an average read depth; ix) a normalized read depth; x) a depth of reads with mapq0 reads; xi) a depth of reads without mapq0 reads; xii) indel statistics (e.g., a polymerase chain reaction (PCR) curve; and / or xiii) a Hidden Markov Model (HMM).xiv) secondary alignment metrics indicating the probability that the secondary genotype call is correct; xv) base context metrics indicating nucleotide context information surrounding the genotype call; xvi) neighborhood call metrics indicating the proximity (e.g., adjacent or within a threshold separation) of the genotype call; xvii) joint detection metrics indicating the probability of detecting a joint corresponding to two or more overlapping genotype calls; and / or xviii) read filter metrics indicating mapping quality, base quality, or other quality metrics or threshold quality metrics or other metrics for filtering out other poor genotype calls. The call integration system 106 generates sequencing metrics 412a and 412b for call model generation from internal (e.g., unique and model-specific) variables reflecting interacting processing pathways, corner cases, and difficult predictions / decisions.

[0117] As indicated above, in some cases, the call integration system 106 determines the FRD score according to the method described in U.S. patent application Ser. No. 16 / 280,022, filed February 19, 2019, to Eric Jon Ojard, entitled "System and Method for Correlated Error Event Mitigation for Variant Calling," which is incorporated herein by reference in its entirety. In certain implementations, the call integration system 106 also (or alternatively) determines the BQD score, FRD score, HMM statistics, and / or other variant calling metrics according to the methods described in U.S. patent application Ser. Nos. 17 / 165,828, 15 / 643,381, and 14 / 811,836, which are incorporated herein by reference in their entireties.

[0118] As shown in FIG. 4B, the sequencing metrics 412a and 412b for call model generation include, but are not limited to, variant calling metrics extracted via the variant calling components of instances of call generation models 410a and 410b. Additionally or alternatively to the above examples of sequencing metrics 412a and 412b for generating a call model, in some cases, the call integration system 106 may process data on i) the number of samples in a population, ii) the number of reads processed to generate a genotype call, the number of variants (e.g., SNPs and indels), iii) the number of biallelic sites (e.g., genomic coordinates containing two observed alleles), iv) the number of multiallelic sites (e.g., the number of sites in a variant call file containing three or more observed alleles), v) the number of SNPs, vi) the number of different types of indels (e.g., homozygous insertions, heterozygous insertions, and heterozygous deletions), vii) the total number of heterozygous indels (e.g., insertions + deletions, insertions + SNPs, or deletions + SNPs), viii) the number of de novo SNPs (e.g., meeting a threshold level), x) the number of SNPs in the first chromosome divided by the number of SNPs in the second chromosome; xii) the number of SNP transitions; xiii) the number of SNP transversions; xiv) the number of heterozygous mutations; xv) the number of homozygous mutations; xvi) the ratio between the number of heterozygous mutations and the number of homozygous mutations; xvii) the number of mutations detected in the dbSNP reference file; and / or xviii) the total number of mutations minus the number detected in the dbSNP file.

[0119] Additionally, the call model generation sequencing metrics 412a and 412b may include mapping and alignment sequencing metrics extracted via the mapping and alignment component of the call generation model 410a or 410b. For example, the call integration system 106 may generate the following sequencing metrics: i) the number of total input reads, ii) the number of duplicate mark reads, iii) the number of mate reads with duplicate marks removed, iv) the number of unique reads, v) the number of reads with mate sequences, vi) the number of reads without mate sequences, vii) an index of reads that fail a quality check, viii) an index of mapped reads, ix) the number of unique and mapped reads, x) the number of unmapped reads, xi) the number of singleton reads, xi) the number of unique and mapped reads, xi ... xii) the number of paired reads; xiii) the number of properly paired reads (e.g., when both reads of a pair are mapped and fall within an acceptable range of each other based on the estimated insert length distribution); xiv) the number of mismatched reads (e.g., the number of reads that are not properly paired); xv) the number of paired reads that are mapped to different chromosomes; xvi) the number of paired reads that are mapped to different chromosomes and also have a mapping quality metric of 10 or higher; xvii) indels R1 and Generate or extract (e.g., via metric re-operation) mapping and alignment metrics, including one or more of: xviii) percentage of reads in R1 and R2; xviii) percentage of soft-clipped bases in R1 and R2; xix) number of mismatched bases in R1 and R2; xx) number of bases (e.g., total and / or R1 or R2) with a base quality of at least 30; xxi) number of alignments (e.g., total, secondary, and / or supplemental alignments); xxii) estimated read length; and xxiii) estimated sample contamination.

[0120] C. External Source Sequencing Metrics 4C , the call integration system 106 generates, extracts, or determines externally sourced sequencing metrics 416. In particular, the call integration system 106 determines the externally sourced sequencing metrics 416 from one or more databases external to the call integration system 106, e.g., a sequencing information database 414. For example, the call integration system 106 accesses sequencing metrics that are general or generally applicable to nucleotide sequencing. Note that the call integration system 106 accesses or determines sequencing information for a particular reference sequence (e.g., stored in the sequencing information database 414).

[0121] In some cases, the call integration system 106 may generate metrics such as: i) a mappability metric that indicates the ease or difficulty of mapping a particular nucleotide sequence (or a particular nucleotide read or nucleobase call) to one or more genomic coordinates in the reference genome; ii) a guanine-cytosine content metric that indicates the count (or dropout or average) of guanine-cytosine content in the reference nucleotide sequence (e.g., the reference genome); iii) a replication timing metric that indicates the time required to replicate a particular number of nucleotides from the reference sequence; iv) one or more DNA structure metrics that indicate the DNA structure of the reference sequence (e.g., the reference genome); and v) a metric that indicates the DNA structure across multiple species. Determine external source sequencing metrics 416, including a conservation metric indicating a measure of sequence conservation (e.g., a measure of change relative to the mean); vi) a confidence classification indicating the degree to which nucleobases at one or more genomic coordinates can be accurately determined; vii) a repeat classification indicating a category of repetitive genomic region for one or more genomic coordinates; viii) a cytosine quadruplex metric indicating that one or more genomic coordinates are part of a cytosine quadruplex; ix) a G-quadruplex metric indicating that one or more genomic coordinates are part of a G-quadruplex; and / or x) a homopolymer metric indicating that one or more genomic coordinates are part of a homopolymer in the reference genome.

[0122] In some embodiments, the call integration system 106 determines external source sequencing metrics 416 by analyzing one or more genomic regions of a reference genome that correspond to (or align with) one or more genomic coordinates for the initial genotype call. Many difficult variant calls occur in low-complexity genomic regions of a reference genome. In some cases, these genomic regions are characterized by some combination of long repeat sequences (e.g., more than 50 base pairs), a very large number (e.g., more than 10) of multiple instances of shorter repeat sequences (e.g., 4-8 repeated bases), and in some cases, contain a subset of bases (e.g., A and T, but not C or G). Nucleotide reads that precisely align to such low-complexity genomic regions often have portions or fragments of the nucleotide read that map to more unique sequences adjacent to the repeat heavy chain region. Alternatively, the reference genome or genomic sample may contain several intermediate breaks (e.g., single bases between primary repeat patterns that break the repetitiveness) that aid in the alignment of nucleotide reads to low-complexity genomic regions of the reference genome. However, when combined with SNPs, indels, and sequencing errors, aligning and collecting reads with sufficient evidence to compare reference versus alternative allele support becomes problematic. Thus, in some embodiments, the call integration system 106 monitors external source sequencing metrics 416 (associated with complexity) that may be augmented with read-based sequencing metrics to provide an overall assessment of the likelihood of the presence of a variant (for both Bayesian and machine learning approaches).

[0123] For example, the call consolidation system 106 accesses or determines sequencing information for a particular reference genome (e.g., stored in a sequencing information database 414). In some cases, the call consolidation system 106 determines externally sourced sequencing metrics 416, including tandem repeat lengths in nucleotide bases, of target genomic regions in the reference genome that correspond to candidate regions in the genomic sample. Specifically, the call consolidation system 106 analyzes portions of the reference genome that correspond to variant regions in the genomic sample to identify tandem repeats (e.g., sequences of two or more bases that are repeated multiple times in a head-to-tail manner) and further determines the lengths (e.g., number of base pairs) within the tandem repeats.

[0124] In certain embodiments, the call integration system 106 determines a sequencing metric for the external source in the form of a repetitiveness metric or a homopolymer metric. Indeed, one metric of the likelihood of mismapping (e.g., mismapping resulting in a false positive) that needs to be corrected is based on the repetitiveness of bases in the reference sequence. Accordingly, the call integration system 106 may utilize various sequencing metrics to measure this repetitiveness, including: i) a maximum repeat pattern length, which indicates the maximum length of a sequence of bases that is repeated at least twice across the extent of the candidate region (corresponding to the reference genome); ii) a maximum repeat length percentage, which indicates the percentage of the region (corresponding to the portion of the reference genome) consumed or occupied by the maximum repeat pattern length; and iii) a maximum homopolymer length, which indicates the length of the longest sequence of the same base in the candidate region (corresponding to the portion of the reference genome).

[0125] In addition to or alternative to the repeatability metric, in some cases, the call integration system 106 determines an externally sourced sequencing metric in the form of permutation entropy of nucleotide bases. For example, the call integration system 106 determines a measure of randomness of nucleotide sequences that may predict mapping / alignment accuracy. In some cases, the call integration system 106 determines permutation entropy by determining the entropy over permutations of a given length of nucleotide sequence. For example, the call integration system 106 may determine permutation entropy according to the following formula:

[0126]

number

[0127]

number

[0128]

number

[0129] As described above, the call integration system 106 may further determine sequencing metrics of the external source in a form that identifies the presence or absence of a cytosine quadruplex (C-quadruplex) or a guanine quadruplex (G-quadruplex) in the target genomic region. Specifically, the call integration system 106 determines counts of cytosine calls and guanine calls within the target genomic region of the reference genome that corresponds to the variant region of the genomic sample or genomic region under consideration for the initial variant call. To identify a cytosine quadruplex, the call integration system 106 identifies occurrences (within the target genomic region) of four or more instantiations of three consecutive cytosine bases separated by one or more different nucleotide bases (e.g., a pattern of CCC A CCC A CCC A CCC). Similarly, to identify a G-quadruplex, the call integration system 106 identifies the occurrence (within the target genomic region) of four or more instantiations of three consecutive guanine bases separated by one or more different nucleotide bases (e.g., a pattern of GGG T GGG T GGG T GGG).

[0130] In one or more embodiments, the call integration system 106 identifies C-quadruplexes or G-quadruplexes in which a maximum threshold number of nucleotide bases (e.g., up to seven nucleotide bases) occurs between instantiations of a triple C or triple G. For example, the call integration system 106 identifies GGG TACC GGG TGTACA GGG AAGTCT GGG as a G-quadruplex. It is well known that, in some cases, G-quadruplexes (and C-quadruplexes) pose sequencing challenges. Thus, the call integration system 106 uses the presence of such sequences to adjust the confidence in mapping and alignment of reads and the accuracy of subsequent sequential sequence assembly.

[0131] In certain embodiments, the call integration system 106 determines a data compression metric as part of the external source sequencing metrics 416. In particular, the call integration system 106 uses one or more data compression algorithms to determine a data compression metric that quantifies a measure of sequence randomness. One such data compression algorithm for lossless compression is the Liv-Zemplel-Welch algorithm. Using this algorithm, the call integration system 106 builds a dictionary of unique k-mers starting with a length of 1 and devise an encoding for each entry in the dictionary. The call integration system 106 may use the number of keys in the dictionary for structural variants and flanking regions in the reference genome as the sequencing metric.

[0132] In addition to or alternatively to the external source sequencing metrics 416 described above, in some embodiments, the call integration system 106 determines structural variant sequence alignment metrics as part of the external source sequencing metrics 416. For example, the call integration system 106 uses gapless alignment scoring and Smith-Waterman alignment scoring of the proposed deletion sequence to the left / right flanking genomic regions in the reference. When there are multiple alignments that score above a threshold gapless alignment score and / or a threshold Smith-Waterman alignment score, the genotype call integration machine learning model may treat the variant sequence alignment metric as a metric of a higher likelihood of an incorrect variant call.

[0133] Furthermore, the call integration system 106 may also determine simulated read alignment metrics as external source sequencing metrics. Assuming that the contiguous sequence representing or including the mutation is accurate, even in the case of a heterozygous deletion, theoretically, there should be many nucleotide reads that have good alignment to the contiguous sequence. However, for true-positive cases with low evidence of mutation, reads corresponding to the SV region may have been mapped elsewhere or not mapped at all, resulting in missed reads. Therefore, the call integration system 106 may determine the likelihood of missing reads by simulating reads.

[0134] Specifically, the call integration system 106 selects a segment from the contiguous sequence that is equal in length to the SBS read. The call integration system 106 selects a segment of contiguous sequence that crosses the interchromosomal translocation(s), is equal to the SBS read length, and aligns to the reference sequence in the SV region. When the alignment is ambiguous, the alternative alignment score will be higher and can serve as a possible guide to the expected read depth. The call integration system 106 may further use a segment of contiguous sequence equal to the read length that is symmetric about the interchromosomal translocation to obtain the highest alignment score. The call integration system 106 may further determine an additional offset from this symmetric point to check the alternative alignment score for the extent of overlap.

[0135] In one or more embodiments, the call integration system 106 determines additional or alternative sequencing metrics, including read-based sequencing metrics, call model-generated sequencing metrics, and / or externally sourced sequencing metrics. For example, the call integration system 106 determines the sequencing metrics in the following table, each of the metrics belonging to one or more of the read-based sequencing metrics, call model-generated sequencing metrics, and / or externally sourced sequencing metrics:

[0136]

Table 1-1

[0137]

Table 1-2

[0138] As noted above, in certain described embodiments, the call integration system 106 uses the sequencing metrics described above to generate a set of machine learning predictions for different variants. In particular, the call integration system 106 utilizes a genotype call integration machine learning model to generate genotype probabilities (for SNPs) or variant call classifications (for indels) corresponding to various genomic coordinates. Additionally, the call integration system 106 determines output genotype calls by generating a variant call file (e.g., a merged variant call file) based on the genotype probabilities and / or variant call classifications. According to one or more embodiments, FIGS. 5A-5C illustrate a call integration system 106 generating one or both of genotype probabilities and variant call classifications, generating genotype calls based on such likelihoods and / or classifications, and generating a merged variant call file containing genotype calls based on such likelihoods and / or classifications. For example, Figure 5A illustrates a call integration system 106 that generates genotype probabilities for (biallelic) SNPs based on sequencing metrics corresponding to initial genotype calls from different read types using a genotype call integration machine learning model, according to one or more embodiments. Figure 5B illustrates a call integration system 106 that generates variant call classifications for indels (or variants other than multi-allelic or bi-allelic SNPs) based on sequencing metrics corresponding to initial genotype calls from different read types using a genotype call integration machine learning model, according to one or more embodiments. Figure 5C then illustrates a call integration system 106 that generates a variant call file including output genotype calls based on the genotype probabilities and / or variant call classifications, according to one or more embodiments.

[0139] 5A , the call integration system 106 identifies a genomic coordinate 502. For example, the call integration system 106 identifies the genomic coordinate 502 from nucleobase calls corresponding to a sample nucleotide sequence or based on haplotype data corresponding to the genomic coordinate 502. In some cases, the call integration system 106 identifies the genomic coordinate 502 by (i) one or more nucleobase calls from nucleotide reads that cover the genomic coordinate and (ii) determining that the one or more nucleobase calls meet one or more threshold sequencing metrics (e.g., a Q30 base call quality metric). Additionally or alternatively, in certain embodiments, the call integration system 106 identifies the genomic coordinate 502 from a database that includes a haplotype reference panel that correlates with the particular genomic coordinate. Regardless of the identification method, in some cases, the call integration system 106 identifies the genomic coordinate 502 using a call generation model 503 (e.g., a mutation caller as part of the call generation model).

[0140] 5A , the call integration system 106 also utilizes a call generation model 503 to generate an initial genotype call 505. Specifically, the call integration system 106 utilizes a call generation model 503 (e.g., DRAGEN caller) to generate the initial genotype call 505 and predict the presence (or absence) of a variant (or a particular genotype) at genomic coordinates 502. As described, the call generation model 503 generates the initial genotype call 505 by analyzing or processing sequencing metrics 504 (or a subset of sequencing metrics 504, such as read-based sequencing metrics and external source sequencing metrics). Note that the call generation model 503 also generates some of the sequencing metrics 504 (e.g., call model-generated sequencing metrics) as part of predicting the initial genotype call 505.

[0141] In effect, the call integration system 106 determines sequencing metrics 504 for the genomic coordinates 502. In particular, the call integration system 106 determines sequencing metrics associated with nucleotide reads generated by the call generation model 503 or retrieved from an external source, as described above. Based on the sequencing metrics 504, the call integration system 106 further generates genotype probabilities 508, which together may indicate a measure of confidence or probability that the genomic coordinates 502 contain or exhibit a SNP mutation.

[0142] 5A , the call integration system 106 utilizes a genotype call integration machine learning model 506 to generate genotype probabilities 508. For example, the genotype call integration machine learning model 506 analyzes or processes the sequencing metrics 504 and the initial genotype call 505 as inputs to generate genotype probabilities 508 as outputs, including: i) a first genotype probability 510 that the initial genotype call 505 is a homozygous reference genotype (e.g., “L(0 / 0)@chr5:4”) at the genomic coordinate 502, ii) a second genotype probability 512 that the initial genotype call 505 is a heterozygous variant genotype (e.g., “L(0 / 1)@chr5:4”) at the genomic coordinate 502, and iii) a third genotype probability 514 that the initial genotype call 505 is a homozygous variant genotype (e.g., “L(1 / 1)@chr5:4”) at the genomic coordinate 502.

[0143] As mentioned, the call integration system 106 generates genotype probabilities 508 to predict whether a SNP occurs at a genomic coordinate 502. However, to predict whether an indel occurs at a genomic coordinate, the call integration system 106 generates a different set of machine learning predictions. Specifically, the call integration system 106 generates a variant call classification indicating the presence (or absence) of an indel (or another variant other than a multi-allelic SNP or bi-allelic SNP) at the genomic coordinate of the sample sequence.

[0144] 5B , the call integration system 106 utilizes a genotype call integration machine learning model 520 to generate a variant call classification 522. Specifically, the call integration system 106 utilizes the genotype call integration machine learning model 520 to generate the variant call classification 522 based on sequencing metrics 518 and an initial genotype call 519 associated with the genomic coordinate 516. Indeed, similar to the above discussion regarding generating genotype probabilities for biallelic SNPs, the call integration system 106 similarly determines sequencing metrics 518 associated with the genomic coordinate 516, including read-based sequencing metrics, call model-generated sequencing metrics, and external-source sequencing metrics. For example, the call integration system 106 utilizes a call generation model 517 to analyze a subset of the sequencing metrics 518 (e.g., read-based sequencing metrics and / or external-source sequencing metrics) to determine an initial genotype call 519 (e.g., indicative of a particular genotype or variant at the genomic coordinate 516). In some cases, the call generation model 517 further generates a subset of sequencing metrics 518 (e.g., call model-generated sequencing metrics) associated with the genome coordinates 516 .

[0145] In generating variant call classifications 522 for genomic coordinates 516, call integration system 106 utilizes genotype call integration machine learning model 520. In particular, call integration system 106 utilizes genotype call integration machine learning model 520 to generate: i) a first true positive variant probability 524 indicating the likelihood that an initial genotype call 519 (or initial VCF file) from a first type of nucleotide read (e.g., SBS reads) is a true positive at genomic coordinates 516; ii) a second true positive variant probability 526 indicating the likelihood that an initial genotype call 519 (or initial VCF file) from a second type of nucleotide read (e.g., assembled nucleotide reads) is a true positive at genomic coordinates 516; iii) a first true positive variant probability 528 indicating the likelihood that an initial genotype call 519 (or initial VCF file) from a second type of nucleotide read (e.g., assembled nucleotide reads) is a true positive at genomic coordinates 516; iv) a first zygosity error probability 528 indicating the likelihood that the initial genotype call 519 (or initial VCF file) from a type of nucleotide read will exhibit a genotype-zygosity error at the genomic coordinate 516, iv) a second zygosity error probability 530 indicating the likelihood that the initial genotype call 519 (or initial VCF file) from a second type of nucleotide read will exhibit a genotype-zygosity error at the genomic coordinate 516, and v) a reference probability 532 indicating the likelihood that the initial genotype call 519 at the genomic coordinate 516 is a homozygous reference genotype (or false positive). In some cases, the variant call classifications 522 are mutually exclusive.

[0146] As shown, the first true positive mutation probability 524 is represented by "TP_s." The symbol "TP_s" represents the probability that the input (x) is a true positive mutation in the first mutation call file (e.g., an SBS mutation call file), and "TP_s" may be formulated as P(tp_s|x)), where "s" represents a first type of nucleotide read, such as a "short read" or SBS read, among others. The second true positive mutation probability 526 is represented by "TP_1." The symbol "~TP_s&TP_1" represents the probability that the input (x) is not a true positive in the first mutation call file (e.g., an SBS mutation call file) but is a true positive in the second mutation call file (e.g., an assembled nucleotide read mutation call file), and "~TP_s&TP_1" may be formulated as P(~tp_s&p_1|x)), where "1" represents a "long read" or assembled nucleotide read.

[0147] Meanwhile, the first zygosity error probability 528 is represented by "HH_s." The symbol "~TP_s&TP_l&HH_s" represents the probability that the input (x) is not a true positive in the first variant call file (e.g., SBS variant call file), is not a true positive in the second variant call file (e.g., assembled nucleotide read variant call file), and is a het-hom error in the first variant call file (e.g., SBS variant call file). The second zygosity error probability 530 is represented by "HH_1." The symbols "~TP_s&~TP_l&~HH_s&HH_l" represent the probability that the input (x) is not a true positive in the first variant call file (e.g., the SBS variant call file), not a true positive in the second variant call file (e.g., the assembled nucleotide read variant call file), not a het-hom error in the first variant call file (e.g., the SBS variant call file), and a het-hom error in the second variant call file (e.g., the assembled nucleotide read variant call file). Additionally, the reference probability 532 is represented by "FP," which indicates the probability that the input (x) is a false positive and may be formulated as P(fp|x)).

[0148] To elaborate the first zygosity error probability 528 and the second zygosity error probability 530, the call integration system 106 determines the probability that a predicted genotype at genomic coordinate 516 (e.g., an initial genotype call for different read types) is an incorrect genotype (e.g., a genotype incorrectly identified by call generation model 517) or includes an incorrect allele. Specifically, in some cases, the call integration system 106 determines, based on the first type of nucleotide read or the second type of nucleotide read, the probability that a zygosity error (e.g., a het / hom error) exists at genomic coordinate 516—e.g., where the alternative base is correct but the genotype is incorrect—or the probability that the nucleobase call represents either the incorrect overall genotype or the incorrect allele(s) in the initial genotype call 519. For example, when determining the probability that a zygosity error exists, the call integration system 106 determines the probability that the alternative base call, represented as "1," is correct but the genotype is incorrect, such as the probability of incorrectly determining a 0 / 1 genotype call (e.g., A / T) instead of a correct 1 / 1 genotype call (e.g., T / T) (or vice versa if the correct genotype call is 0 / 1).

[0149] By determining the first zygosity error probability 528 and the second zygosity error probability 530, the call consolidation system 106 can correct the inaccuracies of existing sequencing systems in which the incorrect calls are often indels. In particular, the call consolidation system 106 can generate more accurate genotype calls for genomic coordinates corresponding to indels when the existing sequencing system determines that the genotype call represents an incorrect genotype that represents an incorrect allele resulting from a long insertion or deletion sequence.

[0150] 5B , the call integration system 106 utilizes a genotype call integration machine learning model 520 to generate a first true positive mutation probability 524 and a second true positive mutation probability 526. Specifically, the call integration system 106 generates the first true positive mutation probability 524 from a first type of nucleotide read (e.g., SBS reads) and generates the second true positive mutation probability 526 from a second type of nucleotide read (e.g., assembled nucleotide reads). In some cases, the true positive mutation probability indicates the probability of a correct mutation call genotype at the genomic coordinate 516. For example, the call integration system 106 generates a probability that the initial genotype call 519 for the genomic coordinate 516 determined by the call generation model 517 is correct.

[0151] Continuing with FIG. 5C , in some embodiments, the call integration system 106 utilizes the genotype probabilities 508 and / or the variant call classifications 522 to update one or more data fields associated with the variant call file or variant call file fields (“VCF” fields). For example, the call integration system 106 generates a merged SNP variant call file 536 based on the genotype probabilities 508 and the variant call classifications 522. Indeed, in some cases, the call integration system 106 generates a single merged variant call file that combines data from the genotype probabilities 508 for SNPs and data from the variant call classifications 522 for indels.

[0152] As shown, call integration system 106 generates updated VCF fields 534 that indicate or correspond to updated sequencing metrics for the output genotype calls. Specifically, call integration system 106 generates one set of updated VCF fields for genotype probabilities 508 and another set of updated VCF fields for variant call classifications 522. For illustrative purposes, FIG. 5C shows some example fields within updated VCF fields 534 without separately showing one set of updated VCF fields for genotype probabilities 508 and another set of updated VCF fields for variant call classifications 522. In some cases, call integration system 106 modifies or updates only certain VCF fields and does not update other fields based on genotype probabilities 508 and / or variant call classifications 522.

[0153] In other cases, the call integration system 106 does not update the VCF field. When generating a genotype call, for example, the call integration system 106 does not update certain fields, such as the genotype (GT) field, based on the genotype probability 508 and / or the variant call classification 522. Indeed, in some cases, the call integration system 106 does not modify or update the GT field because there may not be enough information at the genomic coordinates to determine a new or updated genotype.

[0154] To illustrate one embodiment, FIG. 5C depicts the call consolidation system 106 generating an updated VCF field 534 for a 1 / 2 genotype (GT), where cytosine represents the reference base (shown as "Ref:C") at the genomic coordinate for the allele corresponding to the reference genome, adenine represents the first alternative base ("Alt 1:A") at the genomic coordinate for the different allele, and thymine represents the second alternative base ("Alt 2:T") at the genomic coordinate for the yet different allele. However, FIG. 5C only shows examples of possible reference bases and possible alternative bases at the genomic coordinate. The call consolidation system 106 may generate genotype probabilities 508 and variant call classifications 522 and modify the corresponding metrics of the VCF field for various other reference bases and alternative bases at the genomic coordinate.

[0155] As further illustrated in FIG. 5C , the call integration system 106 generates an updated base call quality (QUAL) field. More specifically, the call integration system 106 modifies or updates the base call quality metric based on the genotype probability 508 and / or the variant call classification 522 to indicate the accuracy of the genotype call. As shown, the updated base call quality field indicates a QUAL score of 48 for the variant at the corresponding genomic coordinate. In this example, the updated base call quality metric (e.g., a QUAL score of 48) represents a score for any type of variant at the corresponding genomic coordinate. Additionally, the call integration system 106 generates a modified or updated genotype quality (GQ) field. For example, based on the variant call classification 522, the call integration system 106 generates a modified or updated genotype quality metric that indicates the likelihood or probability that the predicted genotype at the genomic coordinate is accurate. As shown, for example, the updated genotype quality field indicates the genotype quality metric for genotype calls with heterozygous genotypes (e.g., a GQ score of 4 for a 1 / 2 genotype for a multi-allelic genomic coordinate).

[0156] In one or more embodiments, the call integration system 106 also generates or updates a genotype probability field and (optionally) uses the genotype probability field to rank alleles. Specifically, the call integration system 106 generates the updated GT field by ordering candidate genotype calls to genomic coordinates according to their respective probabilities of belonging to the multi-allelic genomic coordinate. For example, the call integration system 106 determines the probability that each diploid genotype is associated with multiple genotypes consisting of a pair of alleles. As another example, the call integration system 106 determines the relative probabilities associated with multiple alleles (e.g., derived from a reference genome, a first alternative allele, and a second alternative allele) belonging to the genomic coordinate.

[0157] In some embodiments, the call integration system 106 also, or alternatively, generates metrics for a PHRED scaled likelihood (PL) field as part of the updated VCF field. For example, the call integration system 106 generates metrics for the PL field, which may indicate, for example, genotypes for a homozygous reference genotype, a heterozygous genotype, and a homozygous alternative genotype (e.g., having PL field nomenclature 9 / 0 / 3, respectively).

[0158] In one or more embodiments, the call integration system 106 generates allele-specific probabilities or likelihoods based on the relative probabilities of the genotype calls corresponding to the alleles from the call generation model and other (non-reference) genotypes identified by the genotype call integration machine learning model. For example, in some embodiments, the call integration system 106 displays a relative probability score for each allele corresponding to each genotype call in a PL field, which shows the normalized PHRED-scaled likelihood of the genotype, and / or a Genotype Probability (GP) field, which shows the log-scaled posterior genotype probability (e.g., log10 scale) of the data (e.g., sequencing metrics) given the called genotypes.

[0159] To motivate modifying a particular VCF field for a SNP, in some cases, the call integration system 106 utilizes a genotype call integration machine learning model to generate genotype probabilities 508 (which sum to 1). In particular, the genotype call integration machine learning model may generate a first genotype probability 510 of 0.1, a second genotype probability 512 of 0.2, and a third genotype probability 514 of 0.7. Based on the genotype probabilities 508 in such an example, the call integration system 106 generates an updated genotype probability field by updating the GT, GP, and PL fields using a combination of information from the genotype call integration machine learning model and the call generation model.

[0160] As further illustrated in FIG. 5C , the call integration system 106 updates the PL fields for the different genotypes (GT). According to a normalized scale of PL scores, a relatively lower score for a genotype (e.g., PL0) represents a relatively higher likelihood that the genotype is present at the genomic coordinate, and a relatively higher score for a genotype (e.g., PL101) represents a relatively lower likelihood that the genotype is present at the genomic coordinate. For example, the call integration system 106 determines a PL score of 111 for the 0 / 0 genotype, a PL score of 52 for the 0 / 1 genotype, and a PL score of 52 for the 1 / 1 genotype. Thus, in FIG. 5C , a PL score of 52 indicates the most likely or selected genotype (e.g., 0 / 1 and 1 / 1 genotypes), and a PL score of 111 indicates the least likely (e.g., 0 / 0 genotype).

[0161] In some cases, the call integration system 106 generates the updated genotype probability field as a ranking of multiple alleles identified via the call generation model (without utilizing a genotype call integration machine learning model). In other cases, the call integration system 106 utilizes a specialized version of a genotype call integration machine learning model that is trained to generate the updated genotype probability field based on the genotype probabilities 508 and / or variant call classifications 522.

[0162] 5C , the call integration system 106 generates or updates a variant call file, such as a merged SNP variant call file 536. For example, the call integration system 106 generates the variant call file from the updated VCF fields 534 corresponding to the genotype probabilities 508 and variant call classifications 522, respectively. Thus, the call integration system 106 generates the merged SNP variant call file 536 for the SNP genotype calls based on the genotype probabilities 508 and / or variant call classifications 522. Indeed, in some embodiments, the call integration system 106 generates a merged variant call file that merges data for SNPs and indels from both the genotype probabilities 508 and variant call classifications 522.

[0163] 5C , the call integration system 106 may generate a merged SNP variant call file 536 to include an updated VCF field 534 that includes base call quality metrics, genotype quality metrics, and / or updated genotype probability fields. For example, the call integration system 106 may select VCF fields from the initial genotype calls generated by the call generation model (e.g., the initial genotype calls for the SBS reads and the initial genotype calls for the assembled nucleotide reads) to include in the merged variant call file. However, in some embodiments, the call integration system 106 does not select fields and instead generates new VCF fields for the merged variant call file by using a genotype call integration machine learning model to process the genotype probabilities 508 and variant call classifications 522.

[0164] As mentioned, in some cases, the call integration system 106 updates only certain fields, while other fields, such as the genotype (GT) field, remain unchanged. For example, the call integration system 106 updates the genotype quality field and the base call quality field. For other data fields, such as the normalized PHRED scaled likelihood (PL) for the genotype and the posterior genotype probability (GP), the call integration system 106 either (i) keeps the field as is, (ii) removes the field, or (iii) updates the field to reflect the GQ for the called genotype and Class 0 output 0 / 0. In some cases, the call integration system 106 maintains the relative probabilities of other genotypes with respect to the called genotype to ensure consistent updates and the highest called genotype. In certain embodiments, by updating only the 0 / 0 and 1 / 2 values, the call integration system 106 maintains the distance of other genotypes from the called genotype. By updating only certain fields, the call integration system may more efficiently generate a (merged) variant call file without regenerating an entirely new variant call file (as done by some conventional systems) and / or without updating all fields (even those not changed by the new prediction).

[0165] Within (or as a result of generating) the merged variant call file, the call consolidation system 106 may include or update one or more output genotype calls (e.g., variant calls) associated with the genomic coordinates, as determined based on the updated VCF field 534. Indeed, to generate the output genotype calls, the call consolidation system 106 may predict nucleobases from candidate alleles at the genomic coordinates (e.g., according to their respective probabilities and metrics indicated by the merged variant call file). Thus, the call consolidation system 106 may generate output SNP and / or indel calls from the merged SNP variant call file 536.

[0166] Because the call integration system 106 generates genotype calls based on multiple read types in a single pipeline (e.g., combining data from each type of read), there are some situations in which nucleotide reads of different types conflict. Indeed, in certain cases, alternate reads for a first type of nucleotide read (e.g., an SBS read) and alternate reads for a second type of nucleotide read (e.g., an assembled nucleotide read) may not match, but the different read types represent different nucleotide bases. In such situations, the call integration system 106 may utilize a machine learning model trained to determine which read data is more accurate between different read types (e.g., by determining which alternate to select between an SBS read and an assembled nucleotide read). In some embodiments, the call integration system 106 resolves conflicts or discrepancies between different read types by automatically selecting the alternate indicated by the SBS read over that indicated by the assembled nucleotide read (or other read type).

[0167] As described above, in certain embodiments, the call integration system 106 trains or tunes the genotype call integration machine learning model by learning model parameters, such as weights and biases, to generate accurate genotype probabilities or accurate variant call classifications. In particular, the call integration system 106 utilizes an iterative training process to adapt or train the genotype call integration machine learning model by adjusting or adding decision trees or learning parameters that result in genotype probabilities (for SNPs) and / or variant call classifications (for indels). FIG. 6 illustrates the call integration system 106 training a genotype call integration machine learning model, according to one or more embodiments. While FIG. 6 shows different instances of the genotype call integration machine learning model to briefly explain the training process, in some embodiments, the call integration system 106 trains and tunes model parameters for one instance or version of the genotype call integration machine learning model and another instance or version of the genotype call integration machine learning model 608 separately from each other. 6, call integration system 106 separately trains genotype call integration machine learning model 606 (e.g., an SNP-specific model) and genotype call integration machine learning model 608 (e.g., an indel-specific model) as different machine learning models based on different ground truth data. Despite being trained as different machine learning models, in some cases, genotype call integration machine learning model 606 and genotype call integration machine learning model 608 each include the same type of machine learning model (e.g., gradient boosting decision tree, deep learning transformer).

[0168] As illustrated in FIG. 6 , the call integration system 106 trains one instance of a genotype call integration machine learning model 606 to generate genotype probabilities for SNPs and trains another instance of a genotype call integration machine learning model 608 to generate variant call classifications for indels. In particular, the call integration system 106 accesses sample sequencing metrics 604 from a database 602 for use as training data. For example, the call integration system 106 accesses the sample sequencing metrics 604, including sample read-based metrics, sample external source sequencing metrics, and sample call model generation sequencing metrics. In particular embodiments, the sample sequencing metrics 604 may be determined, generated, or derived from multiple different genomic samples analyzed or processed by different sequencing devices. Indeed, the call integration system 106 may train the genotype call integration machine learning model 606 and / or the genotype call integration machine learning model 608 using sample sequencing metrics 604 having different dimensions of variability. Specifically, the sample sequencing metric 604 can vary in the coverage or amount of sequencing performed on the sample to obtain the sequencing metric. The sample sequencing metric 604 can also (or alternatively) vary in the library preparation method, the sequencing device used to obtain the sample sequencing metric 604, and the sequencing run quality (e.g., Q30, error rate, and / or %PF for percent pass filter).

[0169] In some cases, the sample sequencing metrics 604 have associated therewith (e.g., stored in database 602) corresponding ground truth variant call files (e.g., as part of ground truth data 620) that indicate the actual VCF fields for the actual genotype calls resulting from the sample sequencing metrics 604. For example, the call integration system 106 utilizes the sample sequencing metrics 604 and ground truth variant call files (e.g., as part of ground truth data 620) from a training dataset generated by the United States Food and Drug Administration (referred to as the PrecisionFDA dataset). In some cases, the sample sequencing metrics 604 include a subset of the sample sequencing metrics for each genotype call in the ground truth variant call file. The ground truth variant call file may have the ground truth genotype calls that correspond to the sample sequencing metrics.

[0170] As mentioned, the call integration system 106 trains a genotype call integration machine learning model 606 for SNP genotype calling. To train the genotype call integration machine learning model 606, the call integration system 106 inputs the sample sequencing metrics 604 and the sample genotype calls 603 (e.g., initial genotype calls generated by a call generation model from the sample sequencing metrics 604) into the genotype call integration machine learning model 606. The genotype call integration machine learning model 606 then generates predicted genotype probabilities 610 from the sample sequencing metrics 604. For example, the genotype call integration machine learning model 606 generates a predicted first genotype probability, a predicted second genotype probability, and a predicted third genotype probability, as described above.

[0171] As part of training the genotype call integration machine learning model 608 for indels, the call integration system 106 inputs the sample sequencing metrics 604 and the sample genotype calls 603 into the genotype call integration machine learning model 608. The genotype call integration machine learning model 608 then generates predicted variant call classifications 612 based on the sample sequencing metrics 604. Specifically, in some embodiments, the genotype call integration machine learning model 608 generates a set of five predicted variant call classifications, including a first true positive variant probability, a second true positive variant probability, a first zygosity error probability, a second true positive zygosity error probability, and a reference probability, as described above.

[0172] Based on the predicted genotype probabilities 610 and / or the predicted variant call classifications 612, the call integration system 106 generates a modified variant call file 614. For example, the call integration system 106 generates a modified variant call file from the predicted genotype probabilities 610 to train the genotype call integration machine learning model 606. Additionally or alternatively, the call integration system 106 generates a modified variant call file from the predicted variant call classifications 612 to train the genotype call integration machine learning model 608.

[0173] 6 , the call integration system 106 performs a comparison 616. Specifically, the call integration system 106 performs the comparison 616 to compare (i) the predicted genotype probabilities 610 to ground truth data 620 (e.g., ground truth genotype probabilities) and / or (ii) the predicted variant call classifications 612 to the ground truth data 620 (e.g., ground truth variant call classifications). In some embodiments, the call integration system 106 utilizes a loss function 618 to perform the comparison 616. For example, the call integration system 106 utilizes a cross-entropy loss function to compare the predicted genotype probabilities 610 to the ground truth genotype probabilities and / or to compare the predicted variant call classifications 612 to the ground truth variant call classifications (e.g., to determine a measure of error or loss therebetween). When the genotype call synthesis machine learning model 606 or 608 is an ensemble of gradient boosted trees, the call synthesis system 106 utilizes a mean squared error loss function (e.g., for regression) and / or a logarithmic loss function (e.g., for classification) as the loss function 618.

[0174] In contrast, in embodiments in which the genotype call integration machine learning model 606 is a neural network, the call integration system 106 may utilize a cross-entropy loss function, an L1 loss function, or a mean squared error loss function as the loss function 618. For example, the call integration system 106 utilizes the loss function 618 to determine the difference between the predicted genotype probabilities 610 and the ground truth genotype probabilities of the ground truth data 620 and / or between the predicted variant call classifications 612 and the ground truth variant call classifications of the ground truth data 620.

[0175] In some embodiments, the call integration system 106 may (i) utilize a call generation model to generate an initial genotype call, and (ii) utilize a genotype call integration machine learning model 606 or 608 to modify data fields corresponding to the variant call file of the initial genotype call to generate a new predicted genotype call. The call integration system 106 outputs such modified or recalibrated values ​​as part of the modified variant call file 614. For example, the call integration system 106 determines recalibrated values ​​for metrics in the modified variant call file 614, including, among others, the call quality metric (QUAL), the genotype metric (GT), and the genotype quality metric (GQ).

[0176] 6, the call integration system 106 performs model fitting 622. In particular, the call integration system 106 adapts the genotype call integration machine learning model 606 or 608 based on the comparison 616. For example, the call integration system 106 performs modifications or adjustments to the parameters (e.g., weights and biases) of the genotype call integration machine learning model 606 or 608 to reduce the loss measure from the loss function 618 and uses the adjusted parameters in subsequent training iterations.

[0177] In the case of gradient-boosted trees, for example, the call integration system 106 trains the genotype call integration machine learning model 606 or 608 based on the gradient of the error determined by the loss function 618. For example, the call integration system 106 solves a (e.g., infinite-dimensional) convex optimization problem while regularizing the objective function to avoid over-fitting. In certain implementations, the call integration system 106 scales the gradient to emphasize correction for under-represented classes (e.g., when there are significantly more true-positive variant calls than false-positive variant calls).

[0178] In some embodiments, as part of solving the optimization problem, the call integration system 106 adds a new weak learner (e.g., a new boosted tree) to the genotype call integration machine learning model 606 or 608 for each successive training iteration. For example, the call integration system 106 finds a feature (e.g., a sequencing metric) that minimizes the loss from the loss function 618 and adds that feature to the tree for the current iteration or starts building a new tree with that feature.

[0179] Additionally or alternatively to gradient boosting decision trees, the call synthesis system 106 trains a logistic regression to learn parameters for generating genotype calls. To avoid over-fitting, the call synthesis system 106 further regularizes based on hyperparameters such as learning rate, stochastic gradient boosting, number of trees, tree depth(s), complexity penalization, and / or L1 / L2 regularization.

[0180] In embodiments where the genotype call integration machine learning model 606 or 608 is a neural network, the call integration system 106 performs model fitting 622 by modifying the internal parameters (e.g., weights) of the genotype call integration machine learning model 606 or 608 to reduce the loss measure of the loss function 618. In effect, the call integration system 106 changes how the genotype call integration machine learning model 606 or 608 analyzes and passes data between layers and neurons by modifying the internal network parameters. Thus, the call integration system 106 improves the accuracy of the genotype call integration machine learning model 606 or 608 over multiple iterations.

[0181] Indeed, in some cases, the call integration system 106 repeats the training process illustrated in FIG. 6 multiple times. For example, the call integration system 106 repeats the iterative learning by selecting a new set of sequencing metrics for the sample genotype calls along with the corresponding ground truth variant call file. The call integration system 106 further generates a new set of predicted genotype probabilities and / or variant call classifications along with the new modified variant call file for each iteration. As described above, the call integration system 106 also compares the genotype calls and / or data fields from the modified variant call file with the calls and / or data fields from the corresponding ground truth variant call file in each iteration. The call integration system 106 also performs model fitting for each iteration. The call integration system 106 repeats this process until the genotype call integration machine learning model 606 or 608 generates predicted genotype probabilities or variant call classifications that result in a genotype call or variant call file that meets a loss threshold measure.

[0182] In some cases, the call integration system 106 uses a validation data set to determine when training is complete. For example, the call integration system 106 determines a loss for the validation data set (e.g., by comparing the validation data to the predicted genotype probabilities 610 and / or the predicted variant call classifications 612). The call integration system 106 may determine that training is complete based on determining that the loss value associated with the validation data set does not decrease (by a threshold amount) for at least a threshold number of iterations (e.g., 10 iterations). In some embodiments, the call integration system 106 may perform training for a threshold number of iterations (e.g., 400 iterations), after which the call integration system 106 determines that training is complete.

[0183] Although not shown in Figure 6, in certain embodiments, call integration system 106 trains and tunes model parameters for a single genotype call integration machine learning model to generate different outputs (e.g., genotype probabilities and variant call classifications) in different training iterations or training epochs. For example, call integration system 106 (i) performs a set of training iterations to train and tune model parameters for genotype call integration machine learning to generate genotype probabilities, and (ii) performs another set of training iterations to train and tune the same genotype call integration machine learning model to generate variant call classifications. However, because two different genotype call integration machine learning models (e.g., a SNP-specific genotype call integration machine learning model and an indel-specific genotype call integration machine learning model) perform better in terms of recovering false-positive and false-negative variants, Figure 6 shows genotype call integration machine learning model 606 and / or genotype call integration machine learning model 608 being trained separately.

[0184] As mentioned, in certain described embodiments, the call integration system 106 utilizes a genotype call integration machine learning model in conjunction with a call generation model to generate genotype calls. In particular, the call integration system 106 utilizes the output of the genotype call integration machine learning model to modify data fields corresponding to the variant call file containing the genotype call(s) originally generated by the call generation model. Figure 7 illustrates the call integration system 106 generating genotype call(s) and modifying fields in the variant call file containing the genotype call(s) and reported metrics based on the output of the genotype call integration machine learning model and the call generation model, according to one or more embodiments.

[0185] As shown in FIG. 7 , the call integration system 106 accesses a sequencing information database 702, a reference sequence 704, and sequence data 708 extrapolated from one or more nucleotide reads (e.g., a first type of nucleotide read and / or a second type of nucleotide read). In practice, the call integration system 106 performs sequencing metric extraction 714 to extract or re-engineer sequencing metrics as described above. For example, the call integration system 106 generates read-based sequencing metrics, external source sequencing metrics, and call model-generated sequencing metrics. In some cases, the call integration system 106 utilizes the mapping and alignment component 710 of the call generation model 724 to determine the mapping and alignment sequencing metrics as described above. Additionally, the call integration system 106 utilizes the variant calling component 712 of the call generation model 724 to generate variant calling metrics as described above. Additionally, the call integration system 106 also determines read-based sequencing metrics and external source sequencing metrics (eg, from the sequencing information database 702 and / or reference sequences 704).

[0186] 7 , the call integration system 106 generates genotype probabilities 716 and / or variant call classifications 718. By analyzing the sequencing metrics, the first genotype call(s) 700a corresponding to the first type of nucleotide reads, and the second genotype call(s) 700b corresponding to the second type of nucleotide reads, the call integration system 106 utilizes a genotype call integration machine learning model 706a, as described herein, to generate genotype probabilities 716 for the SNPs. Note that by analyzing the sequencing metrics, the first genotype call(s) 700a corresponding to the first type of nucleotide reads, and the second genotype call(s) 700b corresponding to the second type of nucleotide reads, the call integration system 106 utilizes a genotype call integration machine learning model 706b, as described herein, to generate variant call classifications 718 for the indels. As described above, the first genotype call(s) 700a corresponding to the first type of nucleotide read and the second genotype call(s) 700b corresponding to the second type of nucleotide read may originate from different read type pipelines.

[0187] In some cases, the genotype call combined machine learning model 706a or 706b is an ensemble of gradient-boosted trees that process sequencing metrics to generate a genotype probability 716 or a variant call classification 718. For example, the genotype call combined machine learning model 706a or 706b includes a series of weak learners, such as nonlinear decision trees, trained with logistic regression to generate the genotype probability 716 or the variant call classification 718. In some cases, the genotype call combined machine learning model 706a or 706b includes metrics within various trees that define how to process sequencing metrics to generate the respective outputs based on the training described above.

[0188] As alluded to above, in some embodiments, the call integration system 106 may utilize the genotype call integration machine learning models 706a and 706b together. For example, the call integration system 106 utilizes the genotype call integration machine learning models 706a and 706b to generate the genotype probabilities 716 and the variant call classifications 718, respectively. For example, the call integration system 106 utilizes two (or more) different genotype call integration machine learning models in parallel, each trained with different random seeds (e.g., for different biases to process the data differently) and / or on different training data for different types of variants, resulting in different prediction outputs.

[0189] In some embodiments, the call integration system 106 further generates a combined set of predictions from the outputs of the different genotype call integration machine learning models 706a and 706b. For example, the call integration system 106 combines (e.g., averages or sums) metrics from the genotype probabilities 716 and variant call classifications 718. In some embodiments, the call integration system 106 determines an average across predictions from the different models and renormalizes the average. In other embodiments, the call integration system 106 learns linear weights and adapts the weights to minimize the overall error or loss. In still other embodiments, the call integration system 106 weights the genotype probabilities and / or variant call classifications of each genotype call integration machine learning model based on the inverse of the average error between the models.

[0190] In one or more implementations, the call integration system 106 further utilizes a meta-model after the genotype call integration machine learning models 706a and 706b. For example, the call integration system 106 generates the genotype probabilities 716 (e.g., genotype probabilities 508) and variant call classifications 718 (e.g., variant call classifications 522) as described above and combines them using a classification combiner machine learning model. Specifically, the call integration system 106 may combine the genotype probabilities and variant call classifications generated from each genotype call integration machine learning model by selecting weights to apply to the variant call classifications generated by each genotype call integration machine learning model. Indeed, in some cases, the call integration system 106 trains the classification combiner machine learning model to determine, select, or predict the respective weights for the genotype call integration machine learning models to result in the highest accuracy or the lowest loss.

[0191] As an example of generating genotype probabilities 716 and / or variant call classifications 718, in some embodiments, the call integration system 106 uses statistics to summarize the mapping quality distribution of reference and alternative supporting reads (e.g., for a comparative mapping quality distribution metric). The call integration system 106 may determine and use the average MAPQ for reads supporting alternative alleles from SBS reads and assembled nucleotide reads. In these or other embodiments, the genotype call integration machine learning model 706a or 706b learns from the data that if the MAPQ of an alternative allele is low (as indicated by the SBS read or assembled nucleotide read) and the depth metric is high relative to the other MAPQs and depth metrics in the distribution, the resulting genotype call is likely to be a false positive. In fact, as the probability of a false positive increases, the MAPQ metric may decrease.

[0192] As a further example, in some cases, the call integration system 106 compares the mapping quality (e.g., MAPQ) associated with the SBS reads and / or assembled nucleotide reads to a mapping quality threshold. For example, the call integration system 106 utilizes a mapping quality threshold, such as a threshold difference between the best alignment score and the next best alignment score. Upon determining that one or more of the mapping qualities for different read types do not meet a threshold, the call integration system 106 adjusts one or more of the genotype probabilities 716 or variant call classifications 718 accordingly (e.g., to select reads with higher MAPQ).

[0193] Additionally (or alternatively), the call integration system 106 may utilize an accumulation of statistical analyses across complex functions (depending on the architecture of the genotype call integration machine learning model 706a or 706b) to determine genotype probabilities 716 and / or variant call classifications 718 to determine how best to fit the data. For example, as described above, the call integration system 106 trains the genotype call integration machine learning model 706a or 706b to minimize loss generated from multiple (different types of) sequencing metrics and determine weights and biases that best fit the data (e.g., where loss is reduced or minimized).

[0194] 7 , in addition to generating genotype probabilities 716 and variant call classifications 718, call integration system 106 performs data field generation 720. More specifically, call integration system 106 generates data fields for one or more variant call files. In some cases, call integration system 106 generates a first variant call file including first genotype call(s) 700a and further generates a second variant call file including second genotype call(s) 700b. As mentioned, call integration system 106 may utilize first genotype call(s) 700a and / or second genotype call(s) 700b to generate predictions such as genotype probabilities 716 and variant call classifications 718. As further shown, the call integration system 106 may use data field generation 720 to generate a merged variant call file 722 (e.g., by combining all or selecting portions of the first variant call file and the second variant call file) to indicate the output genotype call. To generate the merged variant call file 722, the call integration system 106 utilizes the variant caller component 712 of the call generation model 724 to modify or maintain the values ​​of such data fields based on the genotype probabilities 716 and / or variant call classifications 718.

[0195] For example, the call integration system 106 modifies various metrics, such as quality metrics, mapping metrics, or other metrics associated with the genotype calls. As mentioned, in some cases, the call integration system 106 selects metrics associated with the first type of nucleotide read or the second type of nucleotide read, and / or metrics associated with the genotype probability 716 for a SNP and / or the variant call classification 718 for an indel. In other cases, the call integration system 106 generates new metrics from data generated by the call generation model 724 and / or the genotype call integration machine learning model 706a or 706b. In certain embodiments, the genotype calls are represented or defined by a merged variant call file 722 that includes metrics corresponding to data fields, such as call quality metrics corresponding to call quality fields, genotype metrics corresponding to genotype fields, and genotype quality metrics corresponding to genotype quality fields.

[0196] In certain embodiments, the call integration system 106 generates (data fields for) genotype calls utilizing the variant caller component 712 along with genotype probabilities 716 and / or variant call classifications 718. For example, the call integration system 106 generates data fields for various metrics of the genotype call, such as the nucleotide(s) included in the call, the call quality (QUAL), the genotype (GT), the genotype quality (GQ), one or more normalized PHRED scaled likelihoods (PL), and / or the genotype probability (GP), for inclusion in the merged variant call file 722 and utilizing the variant caller component 712.

[0197] In one or more embodiments, the call integration system 106 recalibrates or corrects the genotype calls (or generates new genotype calls) using the genotype probabilities 716 from the genotype call integration machine learning model 706a and / or the variant call classifications 718 from the genotype call integration machine learning model 706b. As described, the call integration system 106 corrects the genotype calls by modifying or recalibrating data fields of one or more metrics associated with the genotype calls (e.g., as included in the merged variant call file 722).

[0198] For example, to update or recalibrate the call quality metric (QUAL) associated with a genotype call, the call integration system 106 determines how each of the genotype probabilities 716 and / or variant call classifications 718 affects or influences the base call quality metric. For example, the call integration system 106 may determine that a high probability of genotype errors results in a lower overall genotype quality and possibly a different overall call quality. As another example, the call integration system 106 may determine that a high probability of false-positive variants results in a lower overall call quality. As yet another example, the call integration system 106 may determine that a high probability of true-positive variants results in a higher overall (variant) call quality. Accordingly, the call integration system 106 updates the genotype along with the genotype quality and call quality associated with the genotype call.

[0199] In one or more implementations, the call integration system 106 generates a combination (e.g., a weighted combination or average) of genotype probabilities 716 and / or variant call classifications 718 to recalibrate the call quality metrics. In particular, the call integration system 106 weights various predictions of genotype probabilities 716 and / or variant call classifications 718 according to their respective impact on (variant) call quality. In some cases, the call integration system 106 weights each genotype probability or variant call classification equally, while in other cases, the call integration system 106 determines different weights for each. In either case, the call integration system 106 determines a weighted combination or weighted average of genotype probabilities 716 and variant call classifications 718 to recalibrate (increase or decrease) the call quality metrics of the genotype call (e.g., the initial variant call).

[0200] To update or recalibrate genotype metrics associated with genotype calls (e.g., in the GT field of merged variant call file 722), call integration system 106 utilizes one or more of genotype probabilities 716 and / or variant call classifications 718. For example, call integration system 106 compares each of the various component predictions to determine which of the genotype probabilities 716 or variant call classifications 718 has the highest probability. In some cases, call integration system 106 utilizes the genotype probability and / or variant call classification with the highest probability to recalibrate the genotype metrics (e.g., from 0 corresponding to the reference base to 1 corresponding to the first alternative supporting read).

[0201] To update or recalibrate the genotype quality metric associated with a genotype call (e.g., in the GQ field of merged variant call file 722), call integration system 106 utilizes one or more of genotype probabilities 716 and / or variant call classifications 718. More specifically, call integration system 106 determines how each of genotype probabilities 716 and / or variant call classifications 718 affects the genotype quality metric. Call integration system 106 recalibrates the genotype quality metric accordingly (e.g., by increasing or decreasing the quality score between 0 and 10, or 0 and 100, or some other scale). For example, if call integration system 106 determines that a higher genotype error probability (generally) indicates a lower genotype quality metric, call integration system 106 reduces the metric accordingly.

[0202] In some cases, the call integration system 106 determines a combination (e.g., a weighted combination or weighted average) of the genotype probabilities 716 and / or variant call classifications 718 to modify the genotype quality metric. For example, the call integration system 106 determines the combined effect that the genotype probabilities 716 and / or variant call classifications 718 have on the genotype quality metric. As another example, the call integration system 106 determines the individual impact that each constituent prediction of the genotype probabilities 716 and / or variant call classifications 718 has on the genotype quality metric and weights each accordingly. The call integration system 106 further recalibrates the genotype quality metric by increasing or decreasing its value based on the indicated probability.

[0203] As described, the call integration system 106 generates output genotype calls from the same set of sequencing metrics (or a subset of sequencing metrics shared between the genotype call integration machine learning models 706a and 706b and the call generation model 724). Indeed, the call integration system 106 may run the genotype call integration machine learning model 706a or 706b in parallel with the call generation model 724 to generate output genotype call metrics, genotype probabilities 716, and variant call classifications 718 for recalibrating the generated metrics.

[0204] In one or more implementations, the call integration system 106 updates or otherwise modifies data fields for the merged variant call file 722 according to a particular algorithm. After modifying such data fields, the call integration system 106 may generate the merged variant call file 722 (e.g., the filtered variant call file) to include metrics that reflect the updated data fields. For example, in some cases, the call integration system 106 updates a QUAL field for each variant based on the probability of a false-positive variant. As noted above, in some cases, QUAL indicates the probability of a certain variant (or other nucleobase call) being present at a given position, as measured by PHRED scaling.

[0205] As alluded to above, in some embodiments, the call integration system 106 increases or decreases the base call quality metric (e.g., Q-score) for a genotype call. Based on the genotype probability 716 and / or variant call classification 718, for example, the call integration system 106 increases the base call quality metric of a genotype call that would not have previously passed the quality filter and determines that the increased base call quality metric now passes the quality filter. In some such cases, the call integration system 106 includes the genotype call with such increased base call quality metric (that passes the quality filter) in the post-filter variant call file. In contrast, in other cases, the call integration system 106 decreases the base call quality metric for a genotype call that would have previously passed the quality filter and determines that the decreased base call quality metric now does not pass the quality filter. In some such cases, the call integration system 106 excludes genotype calls with reduced base call quality metrics (that do not pass the quality filter) from the post-filter variant call file, but includes genotype calls with such reduced base call quality metrics in the pre-filter variant call file.

[0206] For example, the call integration system 106 may remove false-positive variant calls and recover false-negative variant calls by modifying the corresponding base call quality metric. To remove false positives, in some cases, the call integration system 106 reduces the base call quality metric of a genotype call that initially passed a quality filter based on the genotype probability 716 and / or variant call classification 718 from the genotype call integration machine learning models 706a and 706b. Based on determining that the reduced base call quality metric is below a threshold metric (e.g., a Q-score of 3.0 or 10.0), the call integration system 106 determines that the genotype call no longer passes the quality filter. Thus, the call integration system 106 filters out or removes false-positive genotype calls that initially passed the filter by modifying their base call quality metric.

[0207] In addition to removing false-positive variant calls based on changes to base call quality metrics, the call consolidation system 106 may remove false-positive variant calls based on changes to genotypes. To remove false positives, in some cases, the call consolidation system 106 changes the genotype of an initial genotype call that indicates a different nucleobase from the reference base (e.g., GT=1 or 2) to the genotype of an updated genotype call that indicates the same nucleobase as the reference base (e.g., GT=0). Based on the genotype being the same as the reference base, the call consolidation system 106 does not identify the genotype call as a variant and, in some cases, excludes the genotype call data from the merged variant call file 722. For example, the call consolidation system 106 may use a null-data indicator for the genotype call (or specific fields) in the merged variant call file 722. In some cases, the call integration system 106 uses null data indicators when a particular sequencing metric does not apply to a particular variant call or VCF field (e.g., when an SBS-based call uses a different metric than an assembled nucleotide read-based call).

[0208] In generating the merged variant call file 722, in some embodiments, the call integration system 106 determines a first pipeline accuracy likelihood for the first pipeline (e.g., based on the first read type) and a second pipeline accuracy likelihood for the second pipeline (e.g., based on the second read type). Specifically, the call integration system 106 determines a first pipeline accuracy likelihood that the first genotype call (e.g., a genotype call generated based on the SBS reads) is more accurate than the second genotype call (e.g., a genotype call generated based on the assembled nucleotide reads). The call integration system 106 also determines a second pipeline accuracy likelihood that the second genotype call is more accurate than the first genotype call. Indeed, the call integration system 106 may use the genotype call integration machine learning models 706a and / or 706b to determine the likelihood or probability that the first genotype call and / or the second genotype call are more accurate. Based on the pipeline accuracy likelihood(s), the call integration system 106 may also generate an output genotype call (and corresponding field in the merged variant call file 722) from the first genotype call and / or the second genotype call.

[0209] To recover a false negative, the call integration system 106 increases the base call quality metric of the genotype call that initially failed the quality filter. Based on determining that the increased base call quality metric exceeds the threshold metric, the call integration system 106 determines that the genotype call passes the quality filter. Thus, the call integration system 106 recovers the false negative genotype call that was initially filtered out by changing its base call quality metric.

[0210] In addition to recovering false-negative variant calls based on changes to base call quality metrics, the call consolidation system 106 may recover false-negative variant calls based on changes to genotypes. To recover false negatives, in some cases, the call consolidation system 106 changes the genotype of an initial genotype call that indicates the same nucleobase as the reference base (e.g., GT=0) to a different genotype of an updated genotype call that indicates a different nucleobase from the reference base (e.g., GT=1 or 2). Based on the different genotype of the updated genotype call and the passing base call quality metrics, the call consolidation system 106 identifies the genotype call as a variant and includes the genotype call in the merged variant call file 722.

[0211] Indeed, in some implementations, the call integration system 106 operates in a particular order utilizing the call generation model 724 and the genotype call integration machine learning models 706a and 706b. For example, the call integration system 106 generates a FASTQ file by converting a BCL file to a FASTQ. Note that the call integration system 106 then utilizes the mapping and alignment component 710 of the call generation model 724 to map and align nucleobases from the sample nucleotide sequence. In some cases, the call integration system 106 maps and aligns nucleobases of the sample sequence relative to the reference sequence 704 (e.g., a reference genome) and / or various alternative supporting reads.

[0212] As described herein, after mapping and alignment, the call integration system 106 then utilizes the variant caller component 712 of the call generation model 724 to generate an initial genotype call for the sample sequence corresponding to the particular genomic coordinate based on various sequencing metrics. The call integration system 106 also applies the genotype call integration machine learning models 706a and 706b to generate genotype probabilities 716 and variant call classifications 718 from the mapping and aligning, variant calling, and / or sequencing metrics extracted from other sources as described above. Based on the genotype probabilities 716 and / or variant call classifications 718, the call integration system 106 recalibrates the genotype calls (e.g., by modifying various data fields corresponding to particular metrics of nucleobase calls, such as QUAL, GT, GQ, GP, and / or PL), as described above.

[0213] In some cases, the call integration system 106 further applies a quality filter to the genotype calls to determine whether the genotype calls pass the quality filter (e.g., a Q20 or other Q-score hard-pass filter). The call integration system 106 then identifies a subset of the genotype calls that represent variations from the reference base and pass the quality filter. The call integration system 106 further generates a modified or updated variant call file (e.g., merged variant call file 722) that includes the subset of genotype calls and recalibrated metrics for the subset of genotype calls, such as an updated QUAL metric, an updated GT metric, an updated GQ metric, an updated GP metric, and / or an updated PL metric.

[0214] As noted above, in certain described embodiments, the call integration system 106 provides improved accuracy over existing sequencing systems. In particular, the call integration system 106 reduces false-positive and false-negative variant genotype calls compared to existing sequencing systems. Indeed, by utilizing a genotype call integration machine learning model based on the described sequencing metrics, the call integration system 106 further improves upon previous versions of the call generation model that did not utilize a genotype call integration machine learning model (but still outperformed other systems). Figures 8-10B illustrate graphs and tables from experiments demonstrating the improved accuracy of the call integration system 106.

[0215] For example, Figure 8 illustrates the performance of a previous version of a call generation model (e.g., a model that does not utilize a genotype calling integrated machine learning model) when generating variant calls based on the PrecisionFDA dataset. For example, the previous version separately analyzes assembled nucleotide reads and SBS reads to generate independent results for SNPs and indels. The model generates variant calls for comparison with ground truth data (e.g., from a PrecisionFDA dataset such as HG001 v4.2.1) to determine performance according to the number of false positives and false negatives.

[0216] As shown, graph 802 corresponds to table 806, and graph 804 corresponds to table 808. Graph 802 depicts a Receiver Operating Characteristic (ROC) curve corresponding to the data in table 806, where an earlier version of the call generation model (e.g., without a machine learning component) determined variant calls for SNPs based independently on assembled nucleotide reads and SBS reads. Similarly, graph 804 depicts an ROC curve for the data in table 808, where the earlier version determined variant calls for indels based independently on assembled nucleotide reads and SBS reads. While the performance of the earlier system is good in each case (e.g., with relatively few FPs (False Positives) and no FNs (False Negatives) compared to other conventional systems), call integration system 106 may nevertheless improve this performance by reducing false positives and / or false negatives.

[0217] For example, Figure 9A illustrates a table comparing the performance of a previous version of a call generation model with the performance of the call integration system 106. As shown, table 902 shows cumulative false positive and false negative (FP+FN) metrics for a variant calling model (SBS+ML+GRAPH) that uses a single read type (e.g., SBS reads) along with machine learning predictions and graph genomes (e.g., Illumina DRAGEN Graph Reference Genome hg19) to generate variant calls for SNPs and indels. Table 902 also shows results from the call integration system 106 that utilizes a genotype call integration machine learning model (in addition to using specific sequencing metrics and machine learning predictions) to generate variant calls (for SNPs and indels) based on both SBS reads and assembled nucleotide reads.

[0218] As illustrated in FIG. 9A , experimenters generated the results in table 902 by using different models to generate variant calls for the HG002 dataset, a particular set of human genomic data available for a particular genomic sample. In a similar manner, table 904 shows the results of both the previous model and the genotype call integration machine learning model in generating variant calls for the HG003 dataset. As shown, the call integration system 106 with the genotype call integration machine learning model outperforms the previous model, resulting in lower FP+FN metrics in each table and higher F1 scores in each case (e.g., for SNPs and indels in tables 902 and 904). Indeed, by leveraging different read sources for different types of reads, the call integration system 106 may generate more accurate variant calls than systems that cannot process multiple read types.

[0219] Continuing with reference to FIG. 9B , table 906 illustrates results generated by experimenters using the genotype call integration machine learning model to generate SNPs for the HG002 and HG003 datasets. Additionally, table 908 illustrates results generated by experimenters when using the genotype call integration machine learning model to generate indels for the HG002 and HG003 datasets. Indeed, over longer training for the genotype call integration machine learning model, the experimenters demonstrated further accuracy improvements beyond the metrics shown in the previous figures. Compared to the conventional system, the accuracy metrics in FIG. 9B (e.g., in tables 906 and 908) show discernible improvements in the accuracy metrics for the genotype call integration machine learning model, particularly in FN, FP, recall, precision, and F1-measure. Indeed, the accuracy metrics for the genotype call integration machine learning model shown in FIG. 9B are improvements over the accuracy metrics of the genotype call integration machine learning model in FIG. 9A , which is a further improvement over the conventional system that does not use the genotype call integration machine learning model.

[0220] As shown in FIG. 10A , graph 1002 depicts ROC curves for comparing the performance of different variant callers in generating variant calls for SNPs. For example, graph 1002 illustrates ROC curves in which the curve with the largest area under the curve generally performs best. As shown, the call integration system 106 with the genotype call integration machine learning model outperforms the other models, including the SBS+ML+GRAPH model (as reflected in the table of FIG. 9 ), a model that generates variant calls (solely) from assembled nucleotide reads (e.g., without further analysis or machine learning techniques), and a model that generates variant calls (solely) from SBS reads (e.g., without further analysis or machine learning techniques). As shown by graph 1002, the genotype call integration machine learning model has the highest area under the curve and the fewest false positives, outperforming the other models across the dataset tested (e.g., the PrecisionFDA dataset).

[0221] 10B illustrates bar graph 1004, which corresponds to graph 1002. Specifically, bar graph 1004 provides an alternative visualization of the comparison between the genotype calling integrated machine learning model and the SBS+ML+GRAPH model in variant calling of SNPs (e.g., chromosomes 20-22). Indeed, bar graph 1004 shows false negatives and false positives, along with their cumulative totals for each model. As shown, the genotype calling integrated machine learning model produces more accurate variant calls than the SBS+ML+GRAPH model, resulting in fewer false negatives, fewer false positives, and fewer FP+FN overall.

[0222] Referring now to FIG. 11 , this figure illustrates an exemplary flowchart of a series of operations for generating an output genotype call using a genotype call integrated machine learning model according to one or more embodiments. While FIG. 11 illustrates operations according to one embodiment, alternative embodiments may omit, add, reorder, and / or modify any of the operations shown in FIG. 11 . The operations of FIG. 11 may be performed as part of a method. Alternatively, a non-transitory computer-readable storage medium may include instructions that, when executed by one or more processing devices, cause a computing device to perform the operations depicted in FIG. 11 . In a still further embodiment, a system comprises at least one processing device and a non-transitory computer-readable medium comprising instructions that, when executed by one or more processing devices, cause the system to perform the operations of FIG. 11 .

[0223] 11 illustrates a series of operations 1100 for generating output genotype calls using a genotype call integration machine learning model. In particular, the series of operations 1100 includes an operation 1102 for receiving a first genotype call for a first read type and a second genotype call for a second read type. For example, operation 1102 may involve receiving, for one or more genomic coordinates of a genomic sample, a first genotype call corresponding to a first type of nucleotide read of a first threshold number of nucleobases and a second genotype call corresponding to a second type of nucleotide read of a second threshold number of nucleobases. The first type of nucleotide read may include nucleotide reads synthesized from sample library fragments shorter than the first threshold number of nucleobases. The second type of nucleotide read may include an assembled nucleotide read assembled from shorter nucleotide reads to form a continuous sequence satisfying a first threshold number of nucleobases, a circular consensus sequencing (CCS) read satisfying a first threshold number of nucleobases, or a nanopore long read satisfying a first threshold number of nucleobases. The first genotype call may include a first variant call or a first reference call. The second genotype call may include a second variant call or a second reference call. In some cases, the first genotype call or the second genotype call includes a null data indicator.

[0224] 11 , the series of operations 1100 may include an operation 1104 of identifying sequencing metrics. In particular, operation 1104 may include identifying sequencing metrics corresponding to the first genotype call or the second genotype call. For example, operation 1104 involves identifying sequencing metrics corresponding to the first genotype call or the second genotype call by identifying one or more of: a first set of sequencing metrics associated with the first genotype call corresponding to a first type of nucleotide read; a second set of sequencing metrics associated with the second genotype call corresponding to a second type of nucleotide read; or a shared set of sequencing metrics associated with both the first and second genotype calls. In some cases, operation 1104 includes identifying sequencing metrics corresponding to the first genotype call or the second genotype call by determining one or more of read-based sequencing metrics, call model generation sequencing metrics, external source sequencing metrics, or second read type sequencing metrics associated with the second genotype call corresponding to the second type of nucleotide read.

[0225] In one or more embodiments, operation 1104 involves identifying read-based sequencing metrics including one or more of: an allele frequency corresponding to an allele for a first genotype call, an allele for a second genotype call, or a different allele for an alternative genotype call different from the first genotype call and the second genotype call; a coverage depth of a first type of nucleotide read corresponding to the first genotype call or a second type of nucleotide read corresponding to the second genotype call; an average coverage depth of the first type of nucleotide read corresponding to the first genotype call or the second type of nucleotide read corresponding to the second genotype call; a mapping quality metric for the first type of nucleotide read corresponding to the first genotype call or the second type of nucleotide read corresponding to the second genotype call, or a nucleobase composition of one or more nucleotide reads from the first type of nucleotide read or the second type of nucleotide read.

[0226] In certain embodiments, operation 1104 includes identifying sequencing metrics for call model generation, including one or more of a genotype metric, a base call quality metric, a genotype quality metric, a genotype probability metric, a genotype likelihood metric (e.g., a non-PHRED scaled likelihood metric or a PHRED scaled likelihood metric) for the first genotype call determined from the first type of nucleotide reads or the second genotype call determined from the second type of nucleotide reads.

[0227] In these or other embodiments, operation 1104 involves identifying sequencing metrics of the external source, including one or more of: a mappability metric indicating the degree of difficulty with which a nucleotide read will map to one or more genomic coordinates in the reference genome; a guanine-cytosine content metric indicating a count of guanine-cytosine content corresponding to the one or more genomic coordinates in the reference genome; a confidence classification or confidence score indicating the degree to which nucleobases at the one or more genomic coordinates can be accurately determined; a repeat classification indicating a category of repetitive genomic region for the one or more genomic coordinates; a metric that the one or more genomic coordinates are part of a cytosine quadruplex (C-quadruplex) in the reference genome; a metric that the one or more genomic coordinates are part of a guanine quadruplex (G-quadruplex) in the reference genome; or a metric that one or more genomic coordinates are part of a homopolymer in the reference genome.

[0228] Additionally, the series of operations 1100 may include operation 1106, which uses a genotype call integration machine learning model to generate genotype probabilities and / or variant call classifications. In particular, operation 1106 may include utilizing the genotype call integration machine learning model to generate genotype probabilities for genotype calls for one or more genomic coordinates based on sequencing metrics. In some cases, operation 1106 includes utilizing the genotype call integration machine learning model to generate variant call classifications for candidate variant calls at one or more genomic coordinates based on sequencing metrics.

[0229] In one or more embodiments, operation 1106 involves generating genotype probabilities by utilizing a genotype calling integrated machine learning model trained on SNP training data to generate genotype probabilities for one or more candidate single nucleotide polymorphisms (SNPs). In certain embodiments, operation 1106 involves generating a first genotype probability for a genomic sample comprising a homozygous reference genotype at one or more genomic coordinates, generating a second genotype probability for a genomic sample comprising a heterozygous variant genotype at one or more genomic coordinates, and generating a third genotype probability for a genomic sample comprising a homozygous variant genotype at one or more genomic coordinates.

[0230] In certain embodiments, operation 1106 involves generating a variant call classification for one or more candidate insertions or deletions (indels) utilizing a genotype calling integrated machine learning model trained with indel training data. Operation 1106 may involve generating a variant call classification for the candidate variant call by generating one or more of: a first true positive variant probability that a first genotype call constitutes a true positive variant for one or more genomic coordinates; a second true positive variant probability that a second genotype call constitutes a true positive variant for the one or more genomic coordinates; a first zygosity error probability that the first genotype call includes a genotype-zygosity error at the one or more genomic coordinates; a second zygosity error probability that the second genotype call includes a genotype-zygosity error at the one or more genomic coordinates; or a reference probability of a homozygous reference genotype at the one or more genomic coordinates.

[0231] In some embodiments, the series of operations 1108 includes operation 1108 generating an output genotype call from the genotype probabilities and / or the variant call classification. In particular, operation 1108 may involve generating an output genotype call for one or more genomic coordinates of the genomic sample based on the genotype probabilities. In some cases, operation 1108 involves generating an output genotype call for one or more genomic coordinates of the genomic sample based on the variant call classification. In particular embodiments, operation 1108 involves generating an output genotype call indicating the presence or absence of a SNP at one or more genomic coordinates of the genomic sample. In some embodiments, operation 1108 involves generating an output genotype call indicating the presence or absence of an indel at one or more genomic coordinates of the genomic sample. Operation 1108 may include selecting the first genotype call or the second genotype call, or generating a genotype call that is different from the first genotype call and the second genotype call.

[0232] In certain embodiments, operation 1108 includes selecting the first genotype call in place of the second genotype call. Selecting the first genotype call in place of the second genotype call may include selecting a homozygous reference genotype call from the first genotype call in place of a heterozygous variant genotype call or a homozygous variant genotype call from the second genotype call, selecting a heterozygous variant genotype call from the first genotype call in place of a homozygous reference genotype call or a homozygous variant genotype call from the second genotype call, or selecting a homozygous variant genotype call from the first genotype call in place of a heterozygous variant genotype call or a homozygous reference genotype call from the second genotype call.

[0233] In some cases, operation 1108 includes selecting a second genotype call instead of the first genotype call by selecting a homozygous reference genotype call from the second genotype call instead of a heterozygous variant genotype call or a homozygous variant genotype call from the first genotype call, selecting a heterozygous variant genotype call from the second genotype call instead of a homozygous reference genotype call or a homozygous variant genotype call from the first genotype call, or selecting a homozygous variant genotype call from the second genotype call instead of a heterozygous variant genotype call or a homozygous reference genotype call from the first genotype call. Operation 1108 may include selecting the first genotype call or the second genotype call, or generating a genotype call that is different from the first genotype call and the second genotype call.

[0234] In one or more embodiments, the series of operations 1100 includes an operation of modifying the genotype metrics, base call quality metrics, genotype quality metrics, genotype probability metrics, genotype likelihood metrics, or PHRED-scaled genotype likelihood metrics based on the genotype probabilities and / or variant call classifications. In these or other embodiments, the series of operations 1100 includes an operation of generating a variant call file that includes the modified genotype metrics, modified base call quality metrics, modified genotype quality metrics, modified genotype probability metrics, modified genotype likelihood metrics, or modified PHRED-scaled genotype likelihood metrics.

[0235] In certain embodiments, the series of operations 1100 includes receiving a first genotype call by receiving the first genotype call as part of a first variant call file based on a first type of nucleotide reads. In the same or other embodiments, the series of operations 1100 includes receiving a second genotype call by receiving the second genotype call as part of a second variant call file based on a second type of nucleotide reads, and generating a merged variant call file including the first genotype call or the second genotype call.

[0236] In some embodiments, the series of operations 1100 includes an operation of determining that the first genotype call includes a first alternative nucleobase that is different from the second alternative nucleobase of the second genotype call. The series of operations 1100 may also include an operation of utilizing a genotype call integration machine learning model to generate, based on the sequencing metrics, a first pipeline accuracy likelihood that the first genotype call is more accurate than the second genotype call and a second pipeline accuracy likelihood that the second genotype call is more accurate than the first genotype call. Furthermore, the series of operations 1100 may include an operation of generating an output genotype call by selecting the first genotype call or the second genotype call for one or more genomic coordinates of the genomic sample based on the first pipeline accuracy likelihood and the second pipeline accuracy likelihood.

[0237] The series of operations 1100 may include an operation for determining that the first true positive mutation probability does not satisfy a likelihood threshold. Additionally, the series of operations 1100 may include an operation for generating or utilizing a second true positive mutation probability based on determining that the first true positive mutation probability does not satisfy a likelihood threshold.

[0238] The methods described herein can be used in conjunction with various nucleic acid sequencing techniques. Particularly applicable techniques involve attaching nucleic acids to fixed positions within an array so that their relative positions do not change, and repeatedly imaging the array. For example, embodiments in which images are obtained in different color channels corresponding to different labels used to distinguish one nucleobase type from another are particularly applicable. In some embodiments, the process of determining the nucleotide sequence of a target nucleic acid (e.g., a nucleic acid polymer) can be an automated process. A preferred embodiment involves sequencing-by-synthesis (SBS) techniques.

[0239] SBS technology generally involves the enzymatic extension of nascent nucleic acid chain by repeatedly adding nucleotide to template chain.In the traditional method of SBS, a single nucleotide monomer can be provided to target nucleic acid in the presence of polymerase in each delivery.However, in the method described herein, in the presence of polymerase during delivery, two or more kinds of nucleotide monomers can be provided to target nucleic acid.

[0240] SBS can utilize nucleotide monomers with terminator moieties or nucleotide monomers lacking any terminator moiety. Methods utilizing nucleotide monomers lacking terminators include, for example, pyrosequencing and sequencing using γ-phosphate-labeled nucleotides, as described in more detail below. In methods using nucleotide monomers without terminators, the number of nucleotides added in each cycle is generally variable and depends on the template sequence and the mode of nucleotide delivery. In SBS techniques utilizing nucleotide monomers with terminator moieties, the terminators can be effectively irreversible under the sequencing conditions used, as in conventional Sanger sequencing using dideoxynucleotides, or they can be reversible, as in the sequencing method developed by Solexa (now Illumina, Inc.).

[0241] SBS techniques can use nucleotide monomers bearing a label moiety or lacking a label moiety. Thus, incorporation events can be detected based on the properties of the label, such as the fluorescence of the label; the properties of the nucleotide monomer, such as molecular weight or charge; by-products of nucleotide incorporation, such as the release of pyrophosphate; and the like. In embodiments in which two or more different nucleotides are present in a sequencing reagent, the different nucleotides can be distinguishable from one another, or alternatively, two or more different labels can be distinguishable under the detection technique used. For example, different nucleotides present in a sequencing reagent can have different labels, which can be distinguished using appropriate optical systems, as exemplified by the sequencing method developed by Solexa (now Illumina, Inc.).

[0242] A preferred embodiment is pyrosequencing, which detects the release of inorganic pyrophosphate (Ppi) when a specific nucleotide is incorporated into a nascent strand (Ronaghi, M., Karamohamed, S., Pettersson, B., Uhlen, M., and Nyren, P. (1996) "Real-time DNA sequencing using detection of pyrophosphate release." Analytical Biochemistry 242(1), 84-9; Ronaghi, M. (2001) "Pyrosequencing sheds light on DNA sequencing." Genome Res. 11(1), 3-11; Ronaghi, M., Uhlen, M., and Nyren, P. (1998) "A sequencing method based on real-time pyrophosphate." Science 281 (5375), 363; U.S. Patent Nos. 6,210,891, 6,258,568, and 6,274,320, the disclosures of which are incorporated herein by reference in their entireties. In pyrosequencing, released Ppis can be detected by their immediate conversion to adenosine triphosphate (ATP) by ATP sulfurase, and the level of generated ATP is detected via luciferase-generated photons. Nucleic acids to be sequenced can be attached to features in an array, and the array can be imaged to capture chemiluminescent signals generated by nucleotide incorporation into the array features. Images can be obtained after treating the array with specific nucleotide types (e.g., A, T, C, or G). Images obtained after the addition of each nucleotide type differ in terms of which features in the array are detected. These differences in the images reflect the different sequence content of the features on the array. However, the relative positions of each feature remain unchanged within the image. The images may be stored, processed, and analyzed using the methods described herein.For example, images obtained after treating the array with each different nucleotide type can be processed in the same manner as exemplified herein for images obtained from different detection channels for reversible terminator-based sequencing methods.

[0243] In another exemplary type of SBS, cycle sequencing is achieved by stepwise addition of reversible terminator nucleotides containing cleavable or photobleachable dye labels, as described, for example, in International Publication No. 04 / 018497 and U.S. Patent No. 7,057,026, the disclosures of which are incorporated by reference. This approach has been commercialized by Solexa (now Illumina, Inc.) and is also described in International Publication Nos. 91 / 06678 and 07 / 123,744, each of which is incorporated by reference herein. The availability of fluorescently labeled terminators, both of which can be reversed and from which the fluorescent label is cleaved, facilitates efficient cyclic reversible termination (CRT) sequencing. Polymerases can also be co-engineered to efficiently incorporate and extend from these modified nucleotides.

[0244] Preferably, in reversible terminator-based sequencing embodiments, the label does not substantially inhibit extension under SBS reaction conditions. However, the detection label may be removable, for example, by cleavage or degradation. Images can be taken after incorporation of the label into arrayed nucleic acid features. In certain embodiments, each cycle involves simultaneous delivery of four different nucleotide types to the array, with each nucleotide type bearing a spectrally distinct label. Four images can then be obtained, each using a detection channel selective for one of the four different labels. Alternatively, different nucleotide types can be added sequentially, with an image of the array being obtained between each addition step. In such embodiments, each image shows nucleic acid features incorporating a particular type of nucleotide. Because the sequence content of each feature is different, different features may or may not be present in different images. However, the relative positions of the features remain unchanged within the image. Images obtained from such reversible terminator-SBS methods can be stored, processed, and analyzed as described herein. Following the imaging step, the label can be removed, and the reversible terminator moiety can be removed for subsequent cycles of nucleotide addition and detection. Removing the label after detection in a particular cycle and before the subsequent cycle has the advantage of reducing background signals and crosstalk between cycles. Examples of useful labeling and removal methods are described below.

[0245] In certain embodiments, some or all of the nucleotide monomers may contain reversible terminators. In such embodiments, the reversible terminator / cleavable fluorophore may comprise a fluorophore attached to the ribose moiety via a 3' ester bond (Metzker, Genome Res. 15:1767-1776 (2005), incorporated herein by reference). Other approaches separate the terminator chemistry from the cleavage of the fluorescent label (Ruparel et al., Proc Natl Acad Sci USA 102:5932-7 (2005), incorporated herein by reference in its entirety). Ruparel et al. describe the development of reversible terminators that use a small 3' allyl group to block extension but can be easily unblocked by brief treatment with a palladium catalyst. The fluorophore was attached to the group via a photocleavable linker that can be easily cleaved by 30 seconds of exposure to long-wavelength UV light. Therefore, either disulfide reduction or photocleavage can be used as a cleavable linker. Another approach to reversible termination is the use of a natural terminator followed by the placement of a bulky dye on the dNTP. The presence of a charged bulky dye on the dNTP can act as an effective terminator through steric and / or electrostatic hindrance. The presence of one incorporation event prevents further incorporation unless the dye is removed. Cleavage of the dye removes the fluorophore, effectively reversing the terminus. Examples of modified nucleotides are also described in U.S. Patent Nos. 7,427,673 and 7,057,026, the disclosures of which are incorporated herein in their entireties.

[0246] Additional exemplary SBS systems and methods that may be utilized with the methods and systems described herein are described in U.S. Patent Application Publication Nos. 2007 / 0166705, 2006 / 0188901, U.S. Patent No. 7,057,026, U.S. Patent Application Publication Nos. 2006 / 0240439, 2006 / 0281109, WO 05 / 065814, U.S. Patent Application Publication Nos. 2005 / 0100900, WO 06 / 064199, WO 07 / 010,251, U.S. Patent Application Publication Nos. 2012 / 0270305, and 2013 / 0260372, the disclosures of which are incorporated herein by reference in their entireties.

[0247] Some embodiments may utilize detection of four different nucleotides using fewer than four different labels. For example, SBS may be performed using the methods and systems described in the incorporated document, U.S. Patent Application Publication No. 2013 / 0079232. As a first example, pairs of nucleotide types may be detected at the same wavelength but may be distinguished based on differences in intensity for one member of the pair or based on changes to one member of the pair (e.g., via chemical, photochemical, or physical modification) that result in the appearance or disappearance of a distinct signal compared to the signal detected for the other member of the pair. As a second example, three of the four different nucleotide types may be detected under certain conditions, while the fourth nucleotide type may have no detectable label under those conditions or be minimally detected under those conditions (e.g., minimal detection due to background fluorescence, etc.). Incorporation of the first three nucleotide types into a nucleic acid may be determined based on the presence of their corresponding signals, and incorporation of the fourth nucleotide type into a nucleic acid may be determined based on the absence or minimal detection of any signal. As a third example, one nucleotide type may include label(s) that are detected in two different channels, while other nucleotide types are detected in no more than one of the channels. The foregoing three exemplary configurations are not considered mutually exclusive and may be used in various combinations.An exemplary embodiment combining all three examples is a fluorescence-based SBS method that uses a first nucleotide type that is detected in a first channel (e.g., dATP, having a label that is detected in the first channel when excited by a first excitation wavelength), a second nucleotide type that is detected in a second channel (e.g., dCTP, having a label that is detected in the second channel when excited by a second excitation wavelength), a third nucleotide type that is detected in both the first and second channels (e.g., dTTP, having at least one label that is detected in both channels when excited by the first excitation wavelength and / or the second excitation wavelength), and a fourth nucleotide type that is not detected in either channel or minimally lacks a detectable label (e.g., unlabeled dGTP).

[0248] Furthermore, as described in incorporated document U.S. Patent Application Publication No. 2013 / 0079232, sequencing data can be obtained using a single channel. In such so-called single-dye sequencing methods, a first nucleotide type is labeled but the label is removed after the first image is generated, and a second nucleotide type is labeled only after the first image is generated. A third nucleotide type retains its label in both the first and second images, and a fourth nucleotide type remains unlabeled in both images.

[0249] Some embodiments may utilize sequencing by ligation techniques. Such techniques utilize DNA ligase to incorporate oligonucleotides and identify their incorporation. The oligonucleotides typically have different labels that correlate with the identity of the specific nucleotide in the sequence to which the oligonucleotide hybridizes. As with other SBS methods, images can be obtained after treating an array of nucleic acid features with labeled sequencing reagents. Each image shows nucleic acid features that incorporate a specific type of label. Because the sequence content of each feature varies, different features may or may not be present in different images, but the relative positions of the features remain constant within the image. Images obtained from ligation-based sequencing methods can be stored, processed, and analyzed as described herein. Exemplary SBS systems and methods that can be utilized with the methods and systems described herein are described in U.S. Patent Nos. 6,969,488, 6,172,218, and 6,306,597, the disclosures of which are incorporated herein by reference in their entireties.

[0250] Some embodiments may utilize nanopore sequencing (Deamer, DW & Akeson, M., "Nanopores and nucleic acids: prospects for ultrarapid sequencing." Trends Biotechnol. 18, 147-151 (2000); Deamer, D. and D. Branton, "Characterization of nucleic acids by nanopore analysis." Acc. Chem. Res. 35:817-825 (2002); Li, J., M. Gershow, D. Stein, E. Brandin, and J. A. Golovchenko, "DNA molecules and configurations in a solid-state nanopore microscope." Nat. Mater. 2:611-615 (2003), the disclosures of which are incorporated herein by reference in their entireties). In such embodiments, the target nucleic acid passes through a nanopore. The nanopore may be a synthetic pore or a biological membrane protein, such as α-hemolysin. As the target nucleic acid passes through the nanopore, each base pair can be identified by measuring fluctuations in the electrical conductance of the pore (U.S. Pat. No. 7,001,792; Soni, GV, and Meller, "A. Progress toward ultrafast DNA sequencing using solid-state nanopores," Clin. Chem. 53, 1996-2001 (2007); Healy, K., "Nanopore-based single-molecule DNA analysis," Nanomed. 2, 459-481 (2007); Cockroft, SL, Chu, J., Amorin, M., and Ghadiri, MR, "A single-molecule nanopore device detects DNA polymerase activity with single-nucleotide resolution." J. Am Chem. Soc. 130, 818-820 (2008), the disclosures of which are incorporated herein by reference in their entireties).Data obtained from nanopore sequencing can be stored, processed, and analyzed as described herein. Specifically, the data can be processed as images according to the exemplary processing of optical and other images described herein.

[0251] Some embodiments may utilize methods involving real-time monitoring of DNA polymerase activity. Nucleotide incorporation can be detected via fluorescence resonance energy transfer (FRET) interactions between fluorophore-containing polymerases and γ-phosphate-labeled nucleotides, for example, as described in U.S. Patent Nos. 7,329,492 and 7,211,414 (each of which is incorporated herein by reference), or nucleotide incorporation can be detected using zero-mode waveguides, for example, as described in U.S. Patent No. 7,315,019 (each of which is incorporated herein by reference), and fluorescent nucleotide analogs and engineered polymerases, for example, as described in U.S. Patent No. 7,405,281 and U.S. Patent Application Publication No. 2008 / 0108082 (each of which is incorporated herein by reference). Illumination can be restricted to a zeptoliter-scale volume around the surface-tethered polymerase so that incorporation of fluorescently labeled nucleotides can be observed with low background (Levene, MJ et al., "Zero-mode waveguides for single-molecule analysis at high concentrations." Science, 299, 682-686 (2003); Lundquist, PM et al., "Parallel confocal detection of single molecules in real time." Opt. Lett. 33, 1026-1028 (2008); Korlach, J. et al., "Selective aluminum passivation for targeted immobilization of single DNA polymerase molecules in zero-mode waveguide nanostructures." Proc. Natl. Acad. Sci. USA 105, 1176-1181 (2008), the disclosures of which are incorporated herein by reference in their entireties). Images obtained from such methods can be stored, processed, and analyzed as described herein.

[0252] Some SBS embodiments involve detecting protons released upon incorporation of a nucleotide into an extension product. For example, sequencing based on the detection of released protons can use commercially available electrical detectors and related technology from Ion Torrent (Guilford, Connecticut, a subsidiary of Life Technologies), or the sequencing methods and systems described in U.S. Patent Application Publication Nos. 2009 / 0026082 (A1), 2009 / 0127589 (A1), 2010 / 0137143 (A1), or 2010 / 0282617 (A1), each of which is incorporated herein by reference. The methods described herein for amplifying target nucleic acids using kinetic exclusion can be readily adapted to substrates used for proton detection. More specifically, the methods described herein can be used to generate clonal populations of amplicons used for proton detection.

[0253] The SBS methods described above can be advantageously performed in a multiplex format, allowing multiple different target nucleic acids to be manipulated simultaneously. In certain embodiments, the different target nucleic acids can be processed in a common reaction vessel or on the surface of a particular substrate. This allows for convenient delivery of sequencing reagents, removal of unreacted reagents, and detection of incorporation events in a multiplexed manner. In embodiments using surface-bound target nucleic acids, the target nucleic acids can be in an array format. In an array format, the target nucleic acids are typically bound to a surface in a spatially distinguishable manner. The target nucleic acids can be bound by direct covalent binding, binding to beads or other particles, or binding to a surface-bound polymerase or other molecule. Arrays can contain a single copy of the target nucleic acid at each site (also referred to as a feature), or multiple copies with the same sequence can be present at each site or feature. Multiple copies can be generated by amplification methods such as bridge amplification or emulsion PCR, described in more detail below.

[0254] The methods described herein may use arrays having features at any of a variety of densities, including, for example, at least about 10 features / cm, 100 features / cm, 500 features / cm, 1,000 features / cm, 5,000 features / cm, 10,000 features / cm, 50,000 features / cm, 100,000 features / cm, 1,000,000 features / cm, 5,000,000 features / cm, or more.

[0255] An advantage of the methods described herein is that they provide rapid and efficient detection of multiple target nucleic acids in parallel. Accordingly, the present disclosure provides an integrated system that can prepare and detect nucleic acids using techniques well known in the art, such as those exemplified above. Thus, the integrated system of the present disclosure can include fluidic components that can deliver amplification and / or sequencing reagents to one or more immobilized DNA fragments, including components such as pumps, valves, reservoirs, and fluid lines. A flow cell can be configured and / or used in the integrated system for detecting target nucleic acids. Exemplary flow cells are described, for example, in U.S. Patent Application Publication No. 2010 / 0111768 (A1) and U.S. Patent Application No. 13 / 273,666, each of which is incorporated herein by reference. As exemplified for the flow cell, one or more of the fluidic components of the integrated system can be used in the amplification and detection methods. Taking nucleic acid sequencing embodiments as an example, one or more of the fluidic components of an integrated system can be used for the amplification methods described herein and for delivering sequencing reagents in sequencing methods such as those exemplified above. Alternatively, an integrated system can include separate fluidic systems for performing amplification methods and detection methods. Examples of integrated sequencing systems that can both generate amplified nucleic acids and sequence nucleic acids include, but are not limited to, the MiSeq™ platform (Illumina, Inc., San Diego, CA) and the devices described in U.S. Patent Application No. 13 / 273,666, which is incorporated herein by reference.

[0256] The sequencing system described above sequences nucleic acid polymers present in a sample received by the sequencing device. As defined herein, "sample" and its derivatives are used in the broadest sense and include any sample, culture, etc., believed to contain a target. In some embodiments, a sample comprises DNA, RNA, PNA, LNA, chimeric, or hybrid forms of nucleic acids. A sample can include any biological, clinical, surgical, agricultural, air, or water sample containing one or more nucleic acids. The term also includes any isolated nucleic acid sample, such as genomic DNA, fresh-frozen, or formalin-fixed, paraffin-embedded nucleic acid sample. It is also contemplated that the term may be derived from a single individual, a collection of nucleic acid samples from genetically related members, nucleic acid samples from genetically unrelated members, nucleic acid samples from a single individual (matched), such as a tumor sample and a normal tissue sample, or a sample from a single source containing two different forms of genetic material, such as maternal and fetal DNA obtained from a maternal subject, or the presence of contaminating bacterial DNA in a sample containing plant or animal DNA. In some embodiments, the source of nucleic acid material may include nucleic acid obtained from a newborn, for example, as typically used in newborn screening.

[0257] A nucleic acid sample may include high molecular weight material, such as genomic DNA (gDNA). A sample may include low molecular weight material, such as nucleic acid molecules obtained from FFPE or archived DNA samples. In another embodiment, the low molecular weight material includes enzymatically or mechanically fragmented DNA. A sample may include cell-free circulating DNA. In some embodiments, a sample may include nucleic acid molecules obtained from biopsies, tumors, scrapings, swabs, blood, mucus, urine, plasma, semen, hair, laser capture microdissection, surgical resection, and other clinical or laboratory samples. In some embodiments, a sample may be an epidemiological, agricultural, forensic, or pathogenic sample. In some embodiments, a sample may include nucleic acid molecules obtained from animals, such as humans or mammalian sources. In another embodiment, a sample may include nucleic acid molecules obtained from non-mammalian sources, such as plants, bacteria, viruses, or fungi. In some embodiments, the source of the nucleic acid molecules may be an archived or extinct sample or species.

[0258] Additionally, the methods and compositions disclosed herein may be useful for amplifying nucleic acid samples with low-quality nucleic acid molecules, such as degraded and / or fragmented genomic DNA from forensic samples. In one embodiment, a forensic sample may include nucleic acids obtained from a crime scene, from a missing persons DNA database, from a laboratory associated with a forensic investigation, or from a forensic sample obtained by a law enforcement agency, one or more military services, or any such personnel. A nucleic acid sample may be crude DNA, including purified samples or lysates derived from, for example, a buccal swab, paper, cloth, or other substrate that may be impregnated with saliva, blood, or other bodily fluids. Thus, in some embodiments, a nucleic acid sample may contain small amounts of DNA, such as genomic DNA, or fragmented DNA fragments. In some embodiments, a target sequence may be present in one or more bodily fluids, including, but not limited to, blood, sputum, plasma, semen, urine, and serum. In some embodiments, a target sequence may be obtained from a victim's hair, skin, tissue sample, autopsy, or corpse. In some embodiments, nucleic acids comprising one or more target sequences may be obtained from a deceased animal or human. In some embodiments, the target sequences may include nucleic acids obtained from non-human DNA, such as microbial, plant, or entomological DNA. In some embodiments, the target sequences or amplified target sequences are for human identification purposes. In some embodiments, the present disclosure generally relates to methods for identifying features of forensic samples. In some embodiments, the present disclosure generally relates to human identification methods using one or more target-specific primers disclosed herein or one or more target-specific primers designed using the primer design criteria outlined herein. In one embodiment, a forensic sample or human identification sample comprising at least one target sequence may be amplified using any one or more of the target-specific primers disclosed herein or using the primer criteria outlined herein.

[0259] Components of the call integration system 106 may include software, hardware, or both. For example, components of the call integration system 106 may include one or more instructions stored on a computer-readable storage medium and executable by a processing unit of one or more computing devices (e.g., client device 108). When executed by one or more processing units, the computer-executable instructions of the call integration system 106 may cause the computing devices to perform the bubble detection methods described herein. Alternatively, components of the call integration system 106 may include hardware, such as a dedicated processing unit for performing a particular function or group of functions. Additionally or alternatively, components of the call integration system 106 may include a combination of computer-executable instructions and hardware.

[0260] Furthermore, the components of the call integration system 106 that perform the functions described herein with respect to the call integration system 106 may be implemented, for example, as part of a standalone application, as a module of an application, as a plug-in to an application, as one or more library functions that can be called by other applications, and / or as a cloud computing model. Thus, the components of the call integration system 106 may be implemented as part of a standalone application on a personal computing device or a mobile device. Additionally or alternatively, the components of the call integration system 106 may be implemented in any application that provides sequencing services, including, but not limited to, Illumina BaseSpace, Illumina DRAGEN, or Illumina TruSight software. "Illumina," "BaseSpace," "DRAGEN," and "TruSight" are registered trademarks or trademarks of Illumina, Inc. in the United States and / or other countries.

[0261] Embodiments of the present disclosure may include or utilize special-purpose or general-purpose computers, including computer hardware such as, for example, one or more processing units and system memory, as discussed in more detail below. Embodiments within the scope of the present disclosure also include physical and other computer-readable media for carrying or storing computer-executable instructions and / or data structures. In particular, one or more of the processes described herein may be embodied in a non-transitory computer-readable medium and implemented, at least in part, as instructions executable by one or more computing devices (e.g., any of the media content access devices described herein). Generally, a processing unit (e.g., a microprocessor) receives instructions from a non-transitory computer-readable medium (e.g., a memory, etc.) and executes those instructions, thereby performing one or more processes, including one or more of the processes described herein.

[0262] Computer-readable media may be any available media that can be accessed by a general-purpose or special-purpose computer system. Computer-readable media that store computer-executable instructions are non-transitory computer-readable storage media (devices). Computer-readable media that carry computer-executable instructions are transmission media. Thus, by way of example, and not by way of limitation, embodiments of the present disclosure may include at least two distinctly different kinds of computer-readable media: non-transitory computer-readable storage media (devices) and transmission media.

[0263] Non-transitory computer-readable storage media (devices) include RAM (Random Access Memory), ROM (Read Only Memory), EEPROM (Erasable Programmable Read Only Memory), CD-ROM (Compact Disk Read-Only Memory), Solid State Drives (SSD) (e.g., based on RAM), Flash memory, Phase-Change Memory (PCM), other types of memory, other optical disk storage, magnetic disk storage, or other magnetic storage, or any other medium that can be used to store desired program code means in the form of computer-executable instructions or data structures and that can be accessed by a general-purpose or special-purpose computer.

[0264] A "network" is defined as one or more data links that enable the transport of electronic data between computer systems and / or modules and / or other electronic devices. When information is transferred or provided to a computer over a network or another communications connection (either hardwired, wireless, or a combination of hardwired or wireless), the computer properly views the connection as a transmission medium. Transmission media may be used to carry desired program code means in the form of computer-executable instructions or data structures and may include networks and / or data links that may be accessed by a general-purpose or special-purpose computer. Combinations of the above should also be included within the scope of computer-readable media.

[0265] Furthermore, upon reaching various computer system components, program code means in the form of computer-executable instructions or data structures may be automatically transferred from transmission media to non-transitory computer-readable storage media (devices) (or vice versa). For example, computer-executable instructions or data structures received over a network or data link may be buffered in RAM within a network interface module (e.g., a NIC) and then ultimately transferred to computer system RAM and / or to less volatile computer storage media (devices) within the computer system. Thus, it should be understood that non-transitory computer-readable storage media (devices) may be included in computer system components that also (or even primarily) utilize transmission media.

[0266] Computer-executable instructions include, for example, instructions and data that, when executed by a processor, cause a general-purpose computer, a special-purpose computer, or a special-purpose processing device to perform a certain function or group of functions. In some embodiments, computer-executable instructions are executed on a general-purpose computer, transforming the general-purpose computer into a special-purpose computer that implements elements of the present disclosure. Computer-executable instructions may be, for example, binaries, intermediate format instructions such as assembly language, or even source code. While the subject matter has been described in language specific to structural features and / or methodological acts, it should be understood that the subject matter defined in the appended claims is not necessarily limited to the described features or acts described above. Rather, the described features and acts are disclosed as example forms of implementing the claims.

[0267] Those skilled in the art will appreciate that the present disclosure may be implemented in networked computing environments having many types of computer system configurations, including personal computers, desktop computers, laptop computers, message processing devices, handheld devices, multi-processing device systems, microprocessor-based or programmable consumer electronics, network PCs, minicomputers, mainframe computers, cellular phones, PDAs, tablets, pagers, routers, switches, etc. The present disclosure may also be implemented in distributed system environments where tasks are performed by both local and remote computer systems that are linked through a network (either by hardwired data links, wireless data links, or a combination of hardwired and wireless data links). In a distributed system environment, program modules may be located in both local and remote memory storage devices.

[0268] Embodiments of the present disclosure may also be implemented in a cloud computing environment. As used herein, "cloud computing" is defined as a model for enabling on-demand network access to a shared pool of configurable computing resources. For example, cloud computing may be used in markets to provide ubiquitous, convenient, on-demand access to a shared pool of configurable computing resources. The shared pool of configurable computing resources can be quickly configured through virtualization, exposed with low management effort or service provider interaction, and then scaled accordingly.

[0269] Cloud computing models may consist of various characteristics, such as, for example, on-demand self-service, wide area network access, resource pooling, rapid elasticity, and measured service. Cloud computing models may also expose various service models, such as, for example, Software as a Service (SaaS), Platform as a Service (PaaS), and Infrastructure as a Service (IaaS). Cloud computing models may also be deployed using different deployment models, such as private cloud, community cloud, public cloud, and hybrid cloud. In this specification and claims, a "cloud-computing environment" is an environment in which cloud computing is employed.

[0270] FIG. 12 illustrates a block diagram of a computing device 1200 that may be configured to perform one or more of the processes described above. It should be understood that one or more computing devices, such as computing device 1200, may implement call integration system 106 and sequencing system 104. As illustrated by FIG. 12 , computing device 1200 may include a processing unit 1202, memory 1204, storage 1206, an I / O (Input / Output) interface 1208, and a communication interface 1210, which may be communicatively coupled by a communication infrastructure 1212. In particular embodiments, computing device 1200 may include fewer or more components than those shown in FIG. 12 . The following paragraphs describe in more detail the components of computing device 1200 shown in FIG. 12 .

[0271] In one or more embodiments, processing unit 1202 includes hardware for executing instructions, such as those comprising a computer program. By way of example and not limitation, to execute instructions for dynamically modifying a workflow, processing unit 1202 may retrieve (or fetch) instructions from an internal register, an internal cache, memory 1204, or storage device 1206, decode them, and execute them. Memory 1204 may be volatile or non-volatile memory used to store data, metadata, and programs for execution by processing unit(s). Storage device 1206 includes a storage device, such as a hard disk, flash disk drive, or other digital storage device, for storing data or instructions for implementing the methods described herein.

[0272] I / O interface 1208 enables a user to provide input to, receive output from, and otherwise transfer data to and receive data from computing device 1200. I / O interface 1208 may include a mouse, a keypad or keyboard, a touchscreen, a camera, an optical scanner, a network interface, a modem, other well-known I / O devices, or a combination of such I / O interfaces. I / O interface 1208 may include one or more devices for presenting output to a user, including, but not limited to, a graphics engine, a display (e.g., a display screen), one or more output drivers (e.g., a display driver), one or more audio speakers, and one or more audio drivers. In particular embodiments, I / O interface 1208 is configured to provide graphical data to a display for presentation to a user. The graphical data may represent one or more graphical user interfaces and / or any other graphical content that may be useful in a particular implementation.

[0273] Communications interface 1210 may include hardware, software, or both. In any case, communications interface 1210 may provide one or more interfaces for communications (e.g., packet-based communications, etc.) between computing device 1200 and one or more other computing devices or networks. By way of example and not limitation, communications interface 1210 may include a network interface controller (NIC) or network adapter for communicating with an Ethernet or other wired-based network, or a wireless NIC (WNIC) or wireless adapter for communicating with a wireless network such as WI-FI.

[0274] Additionally, communication interface 1210 may facilitate communication with various types of wired or wireless networks. Communication interface 1210 may also facilitate communication using various communication protocols. Communication infrastructure 1212 may also include hardware, software, or both that couple components of computing device 1200 to one another. For example, communication interface 1210 may use one or more networks and / or protocols to enable multiple computing devices connected by a particular infrastructure to communicate with each other to perform one or more aspects of the processes described herein. By way of example, a sequencing process may enable multiple devices (e.g., client device, sequencing device, and server device(s)) to exchange information such as sequencing data and error notifications.

[0275] In the foregoing specification, the present disclosure has been described with reference to certain exemplary embodiments thereof. Various embodiments and aspects of the present disclosure are described with reference to the details discussed herein, and the accompanying drawings illustrate various embodiments. The above description and drawings are illustrative of the present disclosure and should not be construed as limiting the present disclosure. Numerous specific details are set forth to provide a thorough understanding of various embodiments of the present disclosure.

[0276] The present disclosure may be embodied in other specific forms without departing from its spirit or essential characteristics. The described embodiments are to be considered in all respects as illustrative only and not restrictive. For example, methods described herein may be implemented using fewer or more steps / actions, or steps / actions may be performed in a different order. Additionally, steps / actions described herein may be repeated or performed in parallel with each other or with different instances of the same or similar steps / actions. The scope of the present application is therefore indicated by the appended claims, rather than by the foregoing description. All changes that come within the meaning and range of equivalency of the claims are intended to be embraced within their scope. [Explanation of symbols]

[0277] 100 Environment 102 Server device(s) 104 Sequencing System 106 Call Integration System 107 Genotype Call Integration Machine Learning Model 108 Client Device 108 (if executed) client device 110 Sequencing Applications 112 Network 114 Sequencing equipment 116 Local server device 302 nucleotide reads 304 nucleotide reads 402 nucleotide reads 404 Mapping 406 Sequencing Metrics 408 Sequence Data 408 Sequence Data 410 Call Generation Model 412 Sequencing Metrics 414 Sequencing Information Database 416 Sequencing Metrics 502 genome coordinates 503 Call Generation Model 504 Sequencing Metrics 505 Initial genotype calling 506 Genotype Call Integration Machine Learning Model 508 Genotype Probability 510 First Genotype Probability 512 Second Genotype Probability 514 Third Genotype Probability 516 genome coordinates 517 Call Generation Model 518 Sequencing Metrics 519 Initial Genotype Call 520 Genotype Call Integration Machine Learning Model 522 Mutation Call Classification 524 First True Positive Mutation Probability 526 Second True Positive Mutation Probability 528 First Zygosity Error Probability 528 Probability 530 Second zygosity error probability 532 Reference Probability 534 VCF Field 536 SNP variant call files 602 databases 603 Sample Genotype Call 604 Sample Sequencing Metrics 606 or 608 Genotype call integrated machine learning model 606 Genotype Call Integration Machine Learning Model 608 Genotype Call Integration Machine Learning Model 610 Predicted Genotype Probabilities 612 Predictive Mutation Call Classification 614 Mutation Call File 616 comparison 618 Loss Function 620 ground truth data 622 Model Fitting 700 First genotype call(s) 700 Second Genotype Call(s) 702 Sequencing Information Database 704 reference sequences 706 Genotype Call Integration Machine Learning Model 708 Sequence Data 710 Alignment Components 712 Mutant Cola Components 714 Sequencing Metric Extraction 716 Genotype Probability 718 Mutation Call Classification 720 Data Field Generation 722 Mutation Call File 724 Call Generation Model 1200 Computing Devices 1202 Processing equipment 1204 memory 1206 Storage device 1208 O (Input / Output) interface 1208 I / O Interface 1210 Communication Interface 1212 Communications Infrastructure

Claims

1. 1. A system comprising: at least one processing unit; a non-transitory computer-readable medium that, when executed by the at least one processing device, provides the system with: receiving, for one or more genomic coordinates of the genomic sample, a first genotype call corresponding to a first type of nucleotide read for a first threshold number of nucleobases and a second genotype call corresponding to a second type of nucleotide read for a second threshold number of nucleobases; identifying a sequencing metric corresponding to the first genotype call or the second genotype call; utilizing a genotype call integrated machine learning model to generate genotype probabilities for the genotype call for the one or more genomic coordinates based on the sequencing metrics; a non-transitory computer-readable medium storing instructions that cause generating an output genotype call for the one or more genomic coordinates of the genomic sample based on the genotype probabilities.

2. the first type of nucleotide reads include nucleotide reads synthesized from sample library fragments shorter than the first threshold number of nucleobases; the second type of nucleotide reads are: assembled nucleotide reads assembled from shorter nucleotide reads to form a contiguous sequence that satisfies the first threshold number of nucleobases; circular consensus sequencing (CCS) reads that satisfy the first threshold number of nucleobases; or and a nanopore long read that meets the first threshold number of nucleobases.

3. When executed by the at least one processing unit, the system: generating the genotype probabilities for one or more candidate single nucleotide polymorphisms (SNPs) by utilizing the genotype calling integrated machine learning model trained with single nucleotide polymorphism (SNP) training data; 10. The system of claim 1, further storing instructions that cause the output genotype call to be generated indicating the presence or absence of a SNP at the one or more genomic coordinates of the genomic sample.

4. When executed by the at least one processing unit, the system: by selecting the first genotype call or the second genotype call; or 2. The system of claim 1, further storing instructions for generating the output genotype call by generating a different genotype call that is different from the first genotype call and the second genotype call.

5. When executed by the at least one processing unit, the system: generating a first genotype probability for the genomic sample comprising a homozygous reference genotype at the one or more genomic coordinates; generating a second genotype probability for the genomic sample containing a heterozygous variant genotype at the one or more genomic coordinates; and generating a third genotype probability for the genomic sample containing a homozygous variant genotype at the one or more genomic coordinates.

6. 2. The system of claim 1, wherein the first genotype call comprises a first variant call or a first reference call, and the second genotype call comprises a second variant call or a second reference call.

7. 10. The system of claim 1, wherein the first genotype call or the second genotype call includes a null data indicator.

8. When executed by the at least one processing unit, the system: modifying a genotype metric, a base call quality metric, a genotype quality metric, a genotype probability metric, a genotype likelihood metric, or a PHRED-scaled genotype likelihood metric based on the genotype probabilities; 2. The system of claim 1, further storing instructions for generating a variant call file including the modified genotype metrics, the modified base call quality metrics, the modified genotype quality metrics, the modified genotype probability metrics, the modified genotype likelihood metrics, or the modified PHRED scaled genotype likelihood metrics.

9. When executed by the at least one processing unit, the system: selecting a homozygous reference genotype call from the first genotype call in place of a heterozygous variant genotype call or a homozygous variant genotype call from the second genotype call; selecting the heterozygous variant genotype call from the first genotype call in place of the homozygous reference genotype call or the homozygous variant genotype call from the second genotype call; or 2. The system of claim 1, further storing instructions for generating the output genotype call by selecting the homozygous variant genotype call from the first genotype call in place of the heterozygous variant genotype call or the homozygous reference genotype call from the second genotype call, and by selecting the first genotype call in place of the second genotype call.

10. When executed by the at least one processing unit, the system: selecting a homozygous reference genotype call from the second genotype call in place of a heterozygous variant genotype call or a homozygous variant genotype call from the first genotype call; selecting the heterozygous variant genotype call from the second genotype call in place of the homozygous reference genotype call or the homozygous variant genotype call from the first genotype call; or 2. The system of claim 1, further storing instructions for generating the output genotype call by selecting the homozygous variant genotype call from the second genotype call in place of the heterozygous variant genotype call or the homozygous reference genotype call from the first genotype call, thereby selecting the second genotype call in place of the first genotype call.

11. When executed by the at least one processing unit, the system: a first set of sequencing metrics associated with the first genotype call corresponding to the first type of nucleotide read; a second set of sequencing metrics associated with the second genotype call corresponding to the second type of nucleotide read; or 2. The system of claim 1, further storing instructions for identifying the sequencing metrics corresponding to the first genotype call or the second genotype call by identifying one or more of a shared set of sequencing metrics associated with both the first genotype call and the second genotype call.

12. 10. The system of claim 1, further storing instructions that, when executed by the at least one processing device, cause the system to identify the sequencing metrics corresponding to the first genotype call or the second genotype call by determining one or more of read-based sequencing metrics, call model generation sequencing metrics, external source sequencing metrics, or second read type sequencing metrics associated with the second genotype call corresponding to the second type of nucleotide read.

13. When executed by the at least one processing unit, the system: allele frequencies corresponding to different alleles for the first genotype call, the second genotype call, or an alternative genotype call that is different from the first genotype call and the second genotype call; and a coverage depth of the first type of nucleotide read corresponding to the first genotype call or the second type of nucleotide read corresponding to the second genotype call; an average coverage depth of the first type of nucleotide reads corresponding to the first genotype call or the second type of nucleotide reads corresponding to the second genotype call; and a mapping quality metric for the first type of nucleotide reads corresponding to the first genotype call or the second type of nucleotide reads corresponding to the second genotype call; or and a nucleobase composition of one or more nucleotide reads from the first type of nucleotide reads or the second type of nucleotide reads. The system of claim 1, further storing instructions for identifying the sequencing metric corresponding to the first genotype call or the second genotype call by identifying a read-based sequencing metric comprising one or more of:

14. 2. The system of claim 1, further storing instructions that, when executed by the at least one processing device, cause the system to identify sequencing metrics corresponding to the first genotype call or the second genotype call by identifying sequencing metrics for call model generation, including one or more of a genotype metric, a base call quality metric, a genotype quality metric, a genotype probability metric, or a PHRED scaled likelihood metric for the first genotype call determined from the first type of nucleotide reads or the second genotype call determined from the second type of nucleotide reads.

15. When executed by the at least one processing unit, the system: a mappability metric indicating the degree of difficulty with which a nucleotide read maps to the one or more genomic coordinates within a reference genome; and a guanine-cytosine content metric indicating a guanine-cytosine content count corresponding to the one or more genomic coordinates within the reference genome; and a confidence classification or confidence score indicating the degree to which a nucleobase at said one or more genomic coordinates can be accurately determined; and a repeat classification indicating a category of a repetitive genomic region for the one or more genomic coordinates; a metric that the one or more genomic coordinates are part of a cytosine quadruplex (C-quadruplex) within the reference genome; and a metric that the one or more genomic coordinates are part of a Guanine quadruplex (G-quadruplex) within the reference genome; or 10. The system of claim 1, further storing instructions for identifying the sequencing metrics corresponding to the first genotype call or the second genotype call by identifying sequencing metrics of an external source including one or more of: a metric that the one or more genomic coordinates are part of a homopolymer within the reference genome; and

16. When executed by the at least one processing unit, the system: receiving the first genotype call as part of a first variant call file based on the first type of nucleotide reads; receiving the second genotype call as part of a second variant call file based on the second type of nucleotide reads; The system of claim 1 , further storing instructions for generating a merged variant call file including the first genotype call or the second genotype call.

17. When executed by the at least one processing unit, the system: determining that the first genotype call includes a first alternative nucleobase that is different from a second alternative nucleobase of the second genotype call; generating, utilizing the genotype call combined machine learning model and based on the sequencing metrics, a first pipeline accuracy likelihood that the first genotype call is more accurate than the second genotype call and a second pipeline accuracy likelihood that the second genotype call is more accurate than the first genotype call; 2. The system of claim 1, further storing instructions to generate the output genotype call by selecting the first genotype call or the second genotype call for the one or more genomic coordinates of the genomic sample based on the first pipeline accuracy likelihood and the second pipeline accuracy likelihood.

18. 1. A system comprising: at least one processing unit; a non-transitory computer-readable medium that, when executed by the at least one processing device, provides the system with: receiving, for one or more genomic coordinates of the genomic sample, a first genotype call corresponding to a first type of nucleotide read for a first threshold number of nucleobases and a second genotype call corresponding to a second type of nucleotide read for a second threshold number of nucleobases; identifying a sequencing metric corresponding to the first genotype call or the second genotype call; utilizing a genotype call integration machine learning model to generate a variant call classification for the candidate variant call at the one or more genomic coordinates based on the sequencing metrics; a non-transitory computer-readable medium storing instructions that cause generating an output genotype call for the one or more genomic coordinates of the genomic sample based on the variant call classification.

19. When executed by the at least one processing unit, the system: generating the variant call classification, the variant call classification comprising utilizing the genotype calling integrated machine learning model trained with indel training data to generate the variant call classification for one or more candidate insertions or deletions (indels); 20. The system of claim 18, further storing instructions for generating the output genotype call indicating the presence or absence of an indel at the one or more genomic coordinates of the genomic sample.

20. When executed by the at least one processing unit, the system: a first true positive mutation probability that the first genotype call constitutes a true positive mutation for the one or more genomic coordinates; and a second true positive mutation probability that the second genotype call constitutes a true positive mutation for the one or more genomic coordinates; and a first zygosity error probability, wherein the first genotype call comprises a genotype-zygosity error at the one or more genomic coordinates; and a second zygosity error probability, wherein the second genotype call includes a genotype-zygosity error at the one or more genomic coordinates; or and a reference probability of a homozygous reference genotype at the one or more genomic coordinates, to generate the variant call classification for the candidate variant call.

21. When executed by the at least one processing unit, the system: determining that the first true positive mutation probability does not satisfy a likelihood threshold; 21. The system of claim 20, further storing instructions for generating or utilizing the second true positive mutation probability based on determining that the first true positive mutation probability does not satisfy the likelihood threshold.

22. When executed by the at least one processing unit, the system: selecting the first genotype call or the second genotype call; or 20. The system of claim 18, further storing instructions for generating the output genotype call by generating a different genotype call that is different from the first genotype call and the second genotype call.

23. the first type of nucleotide reads include nucleotide reads synthesized from sample library fragments shorter than the first threshold number of nucleobases; the second type of nucleotide reads are: assembled nucleotide reads assembled from shorter nucleotide reads to form a contiguous sequence that satisfies the first threshold number of nucleobases; circular consensus sequencing (CCS) reads that satisfy the first threshold number of nucleobases; or and a nanopore long read that satisfies the first threshold number of nucleobases.

24. 20. The system of claim 18, wherein the first genotype call comprises a first variant call or a first reference call, and the second genotype call comprises a second variant call or a second reference call.

25. 20. The system of claim 18, wherein the first genotype call or the second genotype call includes a null data indicator.

26. When executed by the at least one processing unit, the system: modifying a genotype metric, a base call quality metric, a genotype quality metric, a genotype probability metric, a genotype likelihood metric, or a PHRED-scaled genotype likelihood metric based on the variant call classification; 20. The system of claim 18, further storing instructions for generating a variant call file including the modified genotype metrics, the modified base call quality metrics, the modified genotype quality metrics, the modified genotype probability metrics, the modified genotype likelihood metrics, or the modified PHRED scaled genotype likelihood metrics.

27. When executed by the at least one processing unit, the system: selecting a homozygous reference genotype call from the first genotype call in place of a heterozygous variant genotype call or a homozygous variant genotype call from the second genotype call; selecting the heterozygous variant genotype call from the first genotype call in place of the homozygous reference genotype call or the homozygous variant genotype call from the second genotype call; or 20. The system of claim 18, further storing instructions for generating the output genotype call by selecting the first genotype call in place of the second genotype call by selecting the homozygous variant genotype call from the first genotype call in place of the heterozygous variant genotype call or the homozygous reference genotype call from the second genotype call.

28. When executed by the at least one processing unit, the system: selecting a homozygous reference genotype call from the second genotype call in place of a heterozygous variant genotype call or a homozygous variant genotype call from the first genotype call; selecting the heterozygous variant genotype call from the second genotype call in place of the homozygous reference genotype call or the homozygous variant genotype call from the first genotype call; or 20. The system of claim 18, further storing instructions for generating the output genotype call by selecting the homozygous variant genotype call from the second genotype call in place of the heterozygous variant genotype call or the homozygous reference genotype call from the first genotype call, thereby selecting the second genotype call in place of the first genotype call.

29. When executed by the at least one processing unit, the system: a first set of sequencing metrics associated with the first genotype call corresponding to the first type of nucleotide read; a second set of sequencing metrics associated with the second genotype call corresponding to the second type of nucleotide read; or 20. The system of claim 18, further storing instructions for identifying the sequencing metrics corresponding to the first genotype call or the second genotype call by identifying one or more of a shared set of sequencing metrics associated with both the first genotype call and the second genotype call.

30. 20. The system of claim 18, further storing instructions that, when executed by the at least one processing device, cause the system to identify the sequencing metrics corresponding to the first genotype call or the second genotype call by determining one or more of read-based sequencing metrics, call model generation sequencing metrics, external source sequencing metrics, or second read type sequencing metrics associated with the second genotype call corresponding to the second type of nucleotide read.

31. When executed by the at least one processing unit, the system: allele frequencies corresponding to different alleles for the first genotype call, the second genotype call, or an alternative genotype call that is different from the first genotype call and the second genotype call; and a coverage depth of the first type of nucleotide read corresponding to the first genotype call or the second type of nucleotide read corresponding to the second genotype call; an average coverage depth of the first type of nucleotide reads corresponding to the first genotype call or the second type of nucleotide reads corresponding to the second genotype call; and a mapping quality metric for the first type of nucleotide reads corresponding to the first genotype call or the second type of nucleotide reads corresponding to the second genotype call; or and a nucleobase composition of one or more nucleotide reads from the first type of nucleotide reads or the second type of nucleotide reads.

19. The system of claim 18, further storing instructions for identifying the sequencing metric corresponding to the first genotype call or the second genotype call by identifying a read-based sequencing metric comprising one or more of:

32. 19. The system of claim 18, further storing instructions that, when executed by the at least one processing device, cause the system to identify sequencing metrics corresponding to the first genotype call or the second genotype call by identifying sequencing metrics for call model generation, including one or more of a genotype metric, a base call quality metric, a genotype quality metric, a genotype probability metric, a genotype likelihood metric, or a PHRED scaled likelihood metric for the first genotype call determined from the first type of nucleotide reads or the second genotype call determined from the second type of nucleotide reads.

33. When executed by the at least one processing unit, the system: a mappability metric indicating the degree of difficulty with which a nucleotide read maps to the one or more genomic coordinates within a reference genome; and a guanine-cytosine content metric indicating a guanine-cytosine content count corresponding to the one or more genomic coordinates within the reference genome; and a confidence classification or confidence score indicating the degree to which a nucleobase at said one or more genomic coordinates can be accurately determined; and a repeat classification indicating a category of a repetitive genomic region for the one or more genomic coordinates; a metric that the one or more genomic coordinates are part of a cytosine quadruplex (C-quadruplex) within the reference genome; and a metric that the one or more genomic coordinates are part of a Guanine quadruplex (G-quadruplex) within the reference genome; or and a metric that the one or more genomic coordinates are part of a homopolymer within the reference genome.

34. When executed by the at least one processing unit, the system: receiving the first genotype call as part of a first variant call file based on the first type of nucleotide reads; receiving the second genotype call as part of a second variant call file based on the second type of nucleotide reads; 20. The system of claim 18, further storing instructions for generating a merged variant call file including the first genotype call or the second genotype call.

35. When executed by the at least one processing unit, the system: determining that the first genotype call includes a first alternative nucleobase that is different from a second alternative nucleobase of the second genotype call; generating, utilizing the genotype call combined machine learning model and based on the sequencing metrics, a first pipeline accuracy likelihood that the first genotype call is more accurate than the second genotype call and a second pipeline accuracy likelihood that the second genotype call is more accurate than the first genotype call; 20. The system of claim 18, further storing instructions to generate the output genotype call by selecting the first genotype call or the second genotype call for the one or more genomic coordinates of the genomic sample based on the first pipeline accuracy likelihood and the second pipeline accuracy likelihood.

Images