IP Library › Granted Patent US 12,217,832
Granted Patent B2
US 12,217,832 · App. 18/314,638 · Granted Feb 4, 2025

Deep learning-based variant classifier

Inventors: Ole Schulz-Trieglaff (Cambridge, GB); Anthony James Cox (Cambridge, GB); Kai-How Farh (San Mateo, CA)
Assignees: Illumina, Inc.; Illumina Cambridge Limited
G16B40/20G06F9/3877G06F18/2148G06F18/2431G06N3/04G06N3/045G06N3/084G16B20/00G16B20/20G16B40/00
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 12,217,832
App. No.
18/314,638
Granted
Feb 4, 2025
Kind
B2
Abstract

The technology disclosed directly operates on sequencing data and derives its own feature filters. It processes a plurality of aligned reads that span a target base position. It combines elegant encoding of the reads with a lightweight analysis to produce good recall and precision using lightweight hardware. For instance, one million training examples of target base variant sites with 50 to 100 reads each can be trained on a single GPU card in less than 10 hours with good recall and precision. A single GPU card is desirable because it a computer with a single GPU is inexpensive, almost universally within reach for users looking at genetic data. It is readily available on could-based platforms.

Claims (58)

1. A system comprising:

at least one processor; and

a non-transitory computer readable medium storing a convolutional neural network and instructions that, when executed by the at least one processor, cause the system to:

identify a group of reads aligned with a reference genome and spanning a candidate variant at a target base position;

provide, to the convolutional neural network, an array of input features generated from a text file comprising sequencing data output by a sequencer instrument, the array of input features encoding:

bases from the group of reads in the text file at the target base position,

bases flanking each side of the target base position in the text file, and

corresponding base features for bases within the group of reads; and

generate, based on an analysis of the array of input features by the convolutional neural network, classification scores indicating likelihoods that the candidate variant at the target base position is a variant.

2. The system of claim 1 , further comprising instructions that, when executed by the at least one processor, cause the system to generate the classification scores by generating a first probability that the candidate variant is a true variant and a second probability that the candidate variant is a false variant.

3. The system of claim 1 , further comprising instructions that, when executed by the at least one processor, cause the system to generate the classification scores by generating one or more of a first score that the candidate variant is a homozygous variant, a second score that the candidate variant is a heterozygous variant, a third score that the candidate variant is a non-variant, or a fourth score that the candidate variant is a complex-variant.

4. The system of claim 1 , further comprising instructions that, when executed by the at least one processor, cause the system to:

process the array of input features through one or more layers of the convolutional neural network to generate convolutional output features;

concatenate the convolutional output features with per-variant characterization data corresponding to the candidate variant at the target base position to generate concatenated features; and

generate the classification scores based on an analysis of the concatenated features by one or more layers of the convolutional neural network.

5. The system of claim 4 , wherein the per-variant characterization data corresponding to the candidate variant comprise empirical variant score (EVS) features for the candidate variant.

6. The system of claim 4 , further comprising instructions that, when executed by the at least one processor, cause the system to generate the classification scores based on the analysis of the concatenated features by one or more fully connected layers and a classification layer of the convolutional neural network.

7. The system of claim 4 , further comprising instructions that, when executed by the at least one processor, cause the system to process the array of input features through the one or more layers of the convolutional neural network by processing the array of input features through a one or more of an input layer, a convolution layer, a batch normalization layer, a max pooling layer, or a flattening layer.

8. The system of claim 1 , further comprising instructions that, when executed by the at least one processor, cause the system to provide the array of input features in which each feature corresponds to a base in a read from the group of reads and each feature comprises a plurality of dimensions.

9. The system of claim 1 , wherein the corresponding base features for bases within the group of reads comprise one or more of:

a corresponding reference base in a reference read corresponding to the reference genome;

a base call accuracy score of calling a base in a read of the group of reads;

a strandedness of calling the base;

an insertion count of changes adjoining a position of the base;

a deletion flag at the position of the base; or

a mapping quality score of aligning a corresponding read that contains the base to the reference read.

10. The system of claim 1 , wherein the text file comprises a binary alignment map (BAM) file.

11. A non-transitory computer readable storage medium storing instructions that, when executed by at least one processor, cause a computing system to:

identify a group of reads aligned with a reference genome and spanning a candidate variant at a target base position;

provide, to a convolutional neural network, an array of input features generated from a text file comprising sequencing data output by a sequencer instrument, the array of input features encoding:

bases from the group of reads in the text file at the target base position,

bases flanking each side of the target base position in the text file, and

corresponding base features for bases within the group of reads; and

generate, based on an analysis of the array of input features by the convolutional neural network, classification scores indicating likelihoods that the candidate variant at the target base position is a variant.

12. The non-transitory computer readable storage medium of claim 11 , further storing instructions that, when executed by the at least one processor, cause the computing system to generate the classification scores by generating a first probability that the candidate variant is a true variant and a second probability that the candidate variant is a false variant.

13. The non-transitory computer readable storage medium of claim 11 , further storing instructions that, when executed by the at least one processor, cause the computing system to generate the classification scores by generating one or more of a first score that the candidate variant is a homozygous variant, a second score that the candidate variant is a heterozygous variant, a third score that the candidate variant is a non-variant, or a fourth score that the candidate variant is a complex-variant.

14. The non-transitory computer readable storage medium of claim 11 , further storing instructions that, when executed by the at least one processor, cause the computing system to:

process the array of input features through one or more layers of the convolutional neural network to generate convolutional output features;

concatenate the convolutional output features with per-variant characterization data corresponding to the candidate variant at the target base position to generate concatenated features; and

generate the classification scores based on an analysis of the concatenated features by one or more layers of the convolutional neural network.

15. The non-transitory computer readable storage medium of claim 14 , wherein the per-variant characterization data corresponding to the candidate variant comprise empirical variant score (EVS) features for the candidate variant.

16. The non-transitory computer readable storage medium of claim 14 , further storing instructions that, when executed by the at least one processor, cause the computing system to generate the classification scores based on the analysis of the concatenated features by one or more fully connected layers and a classification layer of the convolutional neural network.

17. A computer-implemented method comprising:

identifying a group of reads aligned with a reference genome and spanning a candidate variant at a target base position;

providing, to a convolutional neural network, an array of input features generated from a text file comprising sequencing data output by a sequencer instrument, the array of input features encoding:

bases from the group of reads in the text file at the target base position,

bases flanking each side of the target base position in the text file, and

corresponding base features for bases within the group of reads; and

generating, based on an analysis of the array of input features by the convolutional neural network, classification scores indicating likelihoods that the candidate variant at the target base position is a variant.

18. The computer-implemented method of claim 17 , wherein generating the classification scores comprises generating a first probability that the candidate variant is a true variant and a second probability that the candidate variant is a false variant.

19. The computer-implemented method of claim 17 , wherein generating the classification scores comprises generating one or more of a first score that the candidate variant is a homozygous variant, a second score that the candidate variant is a heterozygous variant, a third score that the candidate variant is a non-variant, or a fourth score that the candidate variant is a complex-variant.

20. The computer-implemented method of claim 17 , wherein the corresponding base features for bases within the group of reads comprise one or more of:

a corresponding reference base in a reference read corresponding to the reference genome;

a base call accuracy score of calling a base in a read of the group of reads;

a strandedness of calling the base;

an insertion count of changes adjoining a position of the base;

a deletion flag at the position of the base; or

a mapping quality score of aligning a corresponding read that contains the base to the reference read.

Assignments (2)
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded May 18, 2023
From: FARH, KAI-HOW
To: ILLUMINA, INC.
Reel/Frame 063683/0842 →
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded May 18, 2023
From: SCHULZ-TRIEGLAFF, OLE; COX, ANTHONY JAMES
To: ILLUMINA CAMBRIDGE LIMITED
Reel/Frame 063693/0306 →
Continuity (3)
Continuation 16247487 · Jan 14, 2019
Provisional Application 62617552 · Jan 15, 2018
Related Publication 20230386611A1 · Nov 30, 2023
References Cited (128)
US 5641658A · Adams et al. · 1997 [cited by applicant]
US 7057026B2 · Bames et al. · 2006 [cited by applicant]
US 7115400B1 · Adessi et al. · 2006 [cited by applicant]
US 7211414B2 · Hardin et al. · 2007 [cited by applicant]
US 7315019B2 · Turner et al. · 2008 [cited by applicant]
US 7329492B2 · Hardin et al. · 2008 [cited by applicant]
US 7405281B2 · Xu et al. · 2008 [cited by applicant]
US 9115353B2 · Klausing et al. · 2015 [cited by applicant]
US 20020055100A1 · Kawashima et al. · 2002 [cited by applicant]
US 20040002090A1 · Mayer et al. · 2004 [cited by applicant]
US 20040096853A1 · Mayer · 2004 [cited by applicant]
US 20070128624A1 · Gormley et al. · 2007 [cited by applicant]
US 20070166705A1 · Milton et al. · 2007 [cited by applicant]
US 20080009420A1 · Schroth et al. · 2008 [cited by applicant]
US 20080108082A1 · Rank et al. · 2008 [cited by applicant]
US 20080280773A1 · Fedurco et al. · 2008 [cited by applicant]
US 20090088327A1 · Rigatti et al. · 2009 [cited by applicant]
US 20110098193A1 · Kingsmore et al. · 2011 [cited by applicant]
US 20130296175A1 · Rafnar et al. · 2013 [cited by applicant]
US 20140143188A1 · Mackey et al. · 2014 [cited by applicant]
US 20140280327A1 · Pham et al. · 2014 [cited by applicant]
US 20140304270A1 · Torkamani et al. · 2014 [cited by applicant]
US 20160085910A1 · Bruand et al. · 2016 [cited by applicant]
US 20170286594A1 · Reid et al. · 2017 [cited by applicant]
US 20170286830A1 · El-Yaniv · 2017 [cited by applicant]
US 20180137406A1 · Howard · 2018 [cited by examiner]
US 20180163261A1 · Zeigler · 2018 [cited by examiner]
US 20190189242A1 · Angiuoli · 2019 [cited by examiner]
US 20190318806A1 · Wise · 2019 [cited by examiner]
US 20200185055A1 · Dubourg-Felonneau · 2020 [cited by examiner]
US 20200194099A1 · Li · 2020 [cited by examiner]
US 20200251183A1 · Kashefhaghighi · 2020 [cited by examiner]
US 20200302297A1 · Jaganathan · 2020 [cited by examiner]
CA 2894317A1 · 2016 [cited by applicant]
CN 102939304A · 2013 [cited by applicant]
CN 105934514A · 2016 [cited by applicant]
CN 106459939A · 2017 [cited by applicant]
JP 2017520821A · 2019 [cited by applicant]
JP 2017033046A · 2020 [cited by applicant]
WO 9106678A1 · 1991 [cited by applicant]
WO 2004018497A2 · 2004 [cited by applicant]
WO 2005068089A2 · 2005 [cited by applicant]
WO 2007010252A1 · 2007 [cited by applicant]
WO 2007123744A2 · 2007 [cited by applicant]
WO 2008041002A2 · 2008 [cited by applicant]
WO 2014142831A1 · 2014 [cited by applicant]
WO 2016209999A1 · 2016 [cited by applicant]
WO 2017017611A1 · 2017 [cited by applicant]
WO 2017153556A1 · 2017 [cited by applicant]
WO 2017214320A1 · 2017 [cited by applicant]
WO 2018093780A1 · 2018 [cited by applicant]
WO WO2019140402A1 · 2019 [cited by examiner]
WO 2022155147A1 · 2022 [cited by applicant]
Arik et al., Deep Voice: Real-time Neural Text-to-Speech, dated 2017, 17 pages. [cited by applicant]
Bentley et. al., Accurate Whole Human Genome Sequencing using Reversible Terminator Chemistry, Supplemental Information, Nature, dated Nov. 6, 2008, 55 pages, [retrieved on Jul. 21, 2021], retrieved from the internet UR… [cited by applicant]
Chen-et al_Rethinking-Atrous-Convolution-for-Semantic-Image-Segmentation _Dec. 5, 2017_14pgs. [cited by applicant]
Chin_Simple Convolutional Neural Network for Genomic Variant Calling with TensorFlow_2017_8pgs. [cited by applicant]
Chin_VariantNET_GitHub_Jan. 9, 2019_4pgs. [cited by applicant]
Ching-et al_Opportunities-and-obstacles-for-deep-learning-in-biology-and- medicine_May 28, 2017_102p. [cited by applicant]
Chollet_Xception: Deep-Learning-with-Depthwise-Separable-Convolutions_2017_8pgs. [cited by applicant]
Current-Challenges-In-Cardiovascular-Molecular-Diagnostics_2017_130pgs.pdf. [cited by applicant]
Dean-et al_Large-Scale-Distributed-Deep-Networks_Dec. 2012_11pgs. [cited by applicant]
DePristo et al., “DeepVariant: Highly Accurate Genomes With Deep Neural Networks,” Google AI Blog, Dec. 4, 2017, 1 pages. [cited by applicant]
Dijke, “Convolutional Neural Networks for Regulatory Genomics”, Jun. 17, 2017, Universiteit Leiden Opleiding Informatica, Master's Thesis, Leiden Institute of Advanced Computer Science (LIACS), Leiden University, The Ne… [cited by applicant]
Dong-et al_Comparison-and-integration-of-deleteriousness-prediction-methods-for-nonsynonymous-SNVS_Dec. 30, 2014_13pgs. [cited by applicant]
Dunn et al., Pisces: An Accurate and Versatile Variant Caller for Somatic and Germline Next-Generation Sequencing Data, dated Mar. 29, 2018, 3 pages. [cited by applicant]
Dunn Pisces Accurate and Versatile Variant Caller for Somatic Germline Next Generation Sequence Data_2018_3p. [cited by applicant]
Dunn-et al_Pisces-An-Accurate-and-Versatile-Single-Sample-Somatic-and-Germline-Variant-Caller_Aug. 20, 2017_1pg. [cited by applicant]
Friedman_Deep-learning-in-GATK4_Dec. 2017_11pgs. [cited by applicant]
GermlineEmpiricalVariantScore-EVS-Mode | Training_Illumina_GitHub_Jan. 12, 2018_6p. [cited by applicant]
GitHub_DeepVariant_Jan. 9, 2019_6pgs. [cited by applicant]
Goodfellow et al., Chapter 9—Convolutional Networks, Deep Learning, MIT Press, dated 2016, 41 pages. [cited by applicant]
Gu et. al., Recent Advances in Convolutional Neural Networks, dated Jan. 5, 2017, 37 pages. [cited by applicant]
He-et al_Deep-Residual-Learning-for-Image-Recognition_Dec. 10, 2015_12pgs. [cited by applicant]
Howard-et al_Mobilenets:Efficient-Convolutional-Neural-Networks-for-Mobile-Vision-Applications_Apr. 17, 2017_9pgs. [cited by applicant]
Huang et al., Densely Connected Convolutional Networks, dated Aug. 17, 2017, 9 pages. [cited by applicant]
Huang-et al_Speed-accuracy-trade-offs-for-modern-convolution_Apr. 25, 2017_10pgs. [cited by applicant]
Illumina_Complete-Secondary-Analysis-Workflow-for-the-Genome-Analyzer_Oct. 19, 2009_8pgs. [cited by applicant]
Illumina_Quality-Scores-for-Next-Generation-Sequencing_Oct. 31, 2011_2pgs. [cited by applicant]
Illumina_Understanding-Illumina-Quality-Scores_Apr. 23, 2014_2pgs. [cited by applicant]
Illumina, Nirvana Wiki, Github, 3 pages. Retrieved on Mar. 9, 2022. Retrieved from the internet [URL: https://github com/Illumina/Nirvana/wiki ]. [cited by applicant]
Illumina Dragen Server Installation Guide, Document # 100000007500 v00, Jan. 2019, 21p. [cited by applicant]
Ioffe et al., Batch Normalization Accelerating Deep Network Training by Reducing Internal Covariate Shift, dated 2015 11 pages. [cited by applicant]
Kim-et al_Strelka2-Fast-and-accurate-variant-calling-for-clinical-sequencing-applications_Sep. 23, 2017_19pgs. [cited by applicant]
Kothen-Hill_Deep-Learning-Mutation-Prediction-Enable-EarlyStage-Lung-Cancer-Detect-Liquid-Biopsy_2018_24p. [cited by applicant]
Lin_Network-in-Network_Mar. 4, 2014_10Pgs. [cited by applicant]
Lizardi et al., Mutation detection and single-molecule counting using isothermal rolling-circle amplification, Nature Genetics, vol. 19, pp. 225-232, dated Jul. 1998. [cited by applicant]
Illumina, Pisces, Github, retrieved on Mar. 9, 2022, 5 pages. Retrieved from the internet [URL: https://github.com/ lumina/Pisces .]. [cited by applicant]
Methods for Strelka Small Variant Caller_Aug. 31, 2017_43pgs. [cited by applicant]
Molteni_Google-is-Giving-Away-AI-that-can-Build-Your-Genome-Sequence_Dec. 7, 2017_5pg. [cited by applicant]
O'Fallon et al,. A “support vector machine for identification of single-nucleotide polymorphisms from next-generation sequencing data,” Bioinformatics 29, No. 11 (2013), p. 1361-1366. [cited by applicant]
Oord et al., WAVENET: A Generative Model for Raw Audio, dated Sep. 19, 2016, 15 pages. [cited by applicant]
Owano_DeepVariant-Tool-to-call-out-variants-in-sequencing-data-goes-open-source_Dec. 11, 2017_3p. [cited by applicant]
Piqueras, Autoregressive Model Based on a Deep Convolutional Neural Network for Audio Generation, Tampere University of Technology, dated 2016, 58 pages. [cited by applicant]
Poplin-et al_Creating-a-universal-SNP-and-small-indel-variant-caller-with-deep-neural-networks_Dec. 14, 2016_24pgs.pdf. [cited by applicant]
Qin-et al_FD-MobileNet-Improved-MobileNet-w-Fast-Downsampling-Strategy_Feb. 11, 2018_5pgs. [cited by applicant]
Rakocevic et al.. Fast and Accurate Genomic Analyses using Genome Graphs. biorxi. URL Inttps://www.biorxiv.org/Content/biorxiv/early/2018/03/20/194530.full.pdf], dated Mar. 20, 2018, 104 pages. [cited by applicant]
Running DeepVariant_Jan. 9, 2019_18pgs. [cited by applicant]
Sandler-etal_MobileNetV2-Inverted-Residuals-and-Linear-Bottlenecks_Apr. 2, 2018_14pgs. [cited by applicant]
Saunders_Strelk accurate somatic small variant calling from sequenced tumor normal sample pairs_2012_7p. [cited by applicant]
Saunders-etal_Supplementary-Methods-for-Strelka-Accurate-somatic-small-variant-calling_Mar. 28, 2012_4p. [cited by applicant]
Shlens_Train-your-own-image-classifier-with-Inception-in-TensorFlow_Mar. 9, 2016-4p. [cited by applicant]
Sifre_analysis-and-interpretation-of-single-nucleotide-variation-in-human-disease_2012_12pgs. [cited by applicant]
Sifre_Rigid-motion-Scattering-for-Image-Classification_Oct. 6, 2014_128pgs. [cited by applicant]
Sifre_Rotation-Scaling-Deformation-Invariant-Scattering-for-Texture-Discrimination_2013_8pgs. [cited by applicant]
Spinella_SNooPerMachineLearningBasedMethodSomaticVariantIdFromLlowpassNextGenSequencing_2016. [cited by applicant]
Srivastava_Dropout A Simple Way to Prevent Neural Networks fromOverfitting_2014_30p. [cited by applicant]
Srivastava et al., Highway Networks, dated 2015, 6 pages. [cited by applicant]
Strelka2_germline-and-somatic-small-variant-callerillumina_GitHub_Jan. 10, 2019_3p. [cited by applicant]
StrelkaUserGuide_Illumina_GitHub_Jan. 12, 2018_10p. [cited by applicant]
Strom_Current-practices-and-guidelines-for-clinical-next-generation-sequencing-oncology-testing_Jan. 2, 2016-9gs.pdf. [cited by applicant]
Stromberb-et al_Clinical-grade-variant-annotation_5pgs. [cited by applicant]
Stromberg-etal_Nirvana-Clinical-Grade-Variant-Annotator_Aug. 20, 2017_1p. [cited by applicant]
SupplementaryNotesForStrelka2_13p. [cited by applicant]
Szegedy_Rethinking-the-Inception-Architecture-for-Computer-Vision_Dec. 11, 2015_10pgs. [cited by applicant]
Szegedy-et al_Going-deeper-with-convolutions_2015_9pgs. [cited by applicant]
Szegedy-etal_Inception-v4-Inception-ResNet-and-the-Impact-of-Residual-Connections-on-_earning Feb. 23, 2016 12pgs. [cited by applicant]
Takata_ExomeSequencIdentifiesNovelMissenseVariantRRM2BAssociatedWOphthalmoplegia_2011_7pgs. [cited by applicant]
Torracinta_AdaptiveSomaticMutationsCallsWithDeepLearningandSemiSimulatedData_Oct. 4, 2016_13p. [cited by applicant]
Wolterink et al., Dilated Convolutional Neural Networks for Cardiovascular MR Segmentation in Congenital Heart Disease, dated 2017, 9pages. [cited by applicant]
Wu, Introduction to Convolutional Neural Networks, Nanjing University, dated 2017, 31 pages. [cited by applicant]
Xie-etal_Aggregated-Residual-Transformations-for-Deep-Neural-Networks_Apr. 11, 2017_10pgs. [cited by applicant]
Xiong et al., “The human splicing code reveals new insights into the genetic determinants of disease,” Science 347, ho. 6218 (2015), 1254806, 10 pages. [cited by applicant]
Yu et al., Multi-Scale Context Aggregation by Dilated Convolutions, ICLR 2016, dated Apr. 30, 2016, 13 pages. [cited by applicant]
Zhang et al., “DeepSplice: Deep classification of novel splice junctions revealed by RNA-seq.” nl 2016 IEEE international conference on bioinformatics and biomedicine (BIBM), pp. 330-333. IEEE, 2016. [cited by applicant]
Zhang_Google-Taught-an-AI-That-Sorts-Cat-Photos-to-Analyze-DNA_Dec. 7, 2017_6pgs. [cited by applicant]
Zhang_ShuffleNetAn-Extremely-Efficient-Convolutional-Neural-Network-for-Mobile-Devices_Dec. 7, 2017_9p. [cited by applicant]
Ryan Poplin et al., “Creating a universal SNP and small indel variant caller with deep neural networks”, bioRxiv, [online], May 17, 2017, [retrieved on May 23, 2024], Internet <URL: https://web.archive.org/web/201705170… [cited by applicant]