Genotyping by sequencing

JP2023550242A5Active Publication Date: 2024-11-14REGENERON PHARMACEUTICALS INC
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
JP2023520266
Authority / Receiving Office
JP · JP
Patent Type
Applications
Current Assignee / Owner
Priority Date
2020-11-19
Filing Date
2021-11-19
Publication Date
2024-11-14
Estimated Expiration
2041-11-19

AI Technical Summary

Technical Problem

Genotyping techniques, particularly DNA microarrays, are labor-intensive, inefficient, and require separate laboratory processes, limiting expandability and customization, while direct sequencing of target regions is costly and inefficient in covering the entire genome.

Method used

A method for producing nucleic acid probes that select directly observed genetic variants, calculate quality scores, and use imputation to cover the genome efficiently, allowing for sequencing-based genotyping that complements unobserved variants.

Benefits of technology

The method achieves high-fidelity genotyping with 98.9% call rate and 99.7% accuracy, covering 2.6% of the genome with 1.4M variants, reducing costs and processing time compared to traditional methods.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The present disclosure provides methods for producing nucleic acid probes for genotyping by sequencing, methods for genotyping a DNA sample by sequencing using a set of nucleic acid probes, and systems for performing such methods.
Need to check novelty before this filing date? Find Prior Art

Description

[Technical Field]

[0001] The present disclosure is directed, in part, to methods for producing nucleic acid probes for genotyping by sequencing, methods for genotyping a DNA sample by sequencing using a set of nucleic acid probes, and systems for performing such methods. [Background technology]

[0002] Whole genome sequencing involves sequencing an individual's entire genome. While the cost of whole genome sequencing is decreasing, it is still a significant cost. The greater the depth of sequencing, the higher the cost. Different parts of the genome require different levels of attention or concern, so the requirements for deep sequencing vary.

[0003] Rather than sequencing at a fixed, expected depth across the entire genome, it is possible to preselect regions of the genome for sequencing (and thus perform most of the sequencing in these regions). Exome sequencing targets the exons of genes by capturing short stretches of DNA that overlap with the gene's exons and then sequencing the short stretches of DNA. Exons are of great interest due to their functional and clinical significance. By directly sequencing exons, genetic variation can be observed in specific individual samples without reference to other samples. Exome sequencing targets only approximately 1% of the genome, yet returns unbiased, functional, and actionable genetic variation at a significantly lower cost than whole-genome sequencing.

[0004] An alternative to sequencing strategies is to observe genetic variation using DNA microarray technology, which was developed earlier and on a larger scale than sequencing. DNA microarray technology, for example, allows DNA chips to assay hundreds of thousands of specific variants at once. These genetic variants typically represent genetic variation across the entire genome. Genotyping arrays, which measure genetic variation at hundreds of thousands to millions of variable sites in DNA, are driving modern human genetics. The variable sites measured by each array are typically selected to represent common genetic variation in one or more populations of interest. This strategy offers an inexpensive and effective alternative to direct whole-genome sequencing and is currently used to genotype millions of DNA samples each year. The resulting data allows consumer genetics companies to estimate an individual's ancestry and match them to their DNA relatives. This has also facilitated genome-wide association studies (GWAS), genomic risk scores, and Mendelian randomization analyses, which have provided significant insight into the biology of diverse complex traits related to human health and behavior, ranging from cardiovascular and metabolic diseases to psychiatric disorders, and from human behavior to age-related disorders and cancer.

[0005] Traditional strategies for array design focus on a set of known common genetic variants and attempt to identify a subset of these variants that are expected to perform well in multiplex genotyping experiments and adequately represent other known common variants. Typically, each variant is assigned a probe score that measures its expected performance on the array platform. This score summarizes factors such as the presence of other nearby variants, repetitiveness, the proportion of guanine-cytosine (GC) bases in the probe DNA sequence, and the performance of similar probes on previous genotyping arrays. Each of these factors can affect the performance of a genotyping probe targeting a variant. In addition to this probe score summarizing the probe's expected performance, variants are also typically mapped to a list of other common variants they may represent. Variants that represent mutations in other nearby common variants are "proxies" or "surrogates" for these additional variants. These proxy relationships are common between nearby variants in the human genome due to a process known as linkage disequilibrium. Linkage disequilibrium is the result of genetic variants entering a population through mutation or introgression, and then spreading gradually through inheritance and recombination and gene conversion. Mutation, introgression, inheritance, recombination, and gene conversion together often result in predictable combinations of nearby genetic variants, which usually reflect the ancestral chromosomes from which each variant first entered the population.

[0006] Genotyping arrays, such as DNA microarrays, observe only a small subset of variants in individual samples. Selecting a set of directly observed variants for inclusion in a genotyping array ultimately involves selecting a set of directly observed variants with high "probe scores" that can serve as a "proxy" for the majority of all known genetic variants. It is possible to indirectly observe (imput) variants from directly observed variants. This process is called imputation. Imputation is successful because our genetic variation is inherited in such a way that the closer multiple variants are to each other on the same chromosome, the higher the probability that they were inherited from the same ancestor. Imputation methods take into account the inferences about how segments of DNA are inherited and have been shown to produce high-quality results for imputing variants not directly observed. While this strategy yields a list of variants that is representative of common genetic variation in humans, it is also inefficient for techniques that measure multiple genetic variants with a single probe. Another problem with DNA microarray assays is that they are entirely separate processes in the laboratory, requiring numerous replications, leading to experimental inefficiencies. What is needed is a cost-effective experimental strategy that allows for the direct sequencing of desired target regions while retaining the ability to complement variants across the entire genome. Summary of the Invention [Problem to be solved by the invention]

[0007] Genotyping technology has remained largely unchanged for nearly 20 years. Arrays produce high-quality data and consistent results at low cost, but are labor-intensive. Arrays require additional processing and equipment that differs from that used for whole-exome sequencing. Arrays are limited in their scalability and customizability. Efficient processing of millions of samples is required. [Means for solving the problem]

[0008] The present disclosure provides a method for producing nucleic acid probes for genotyping by sequencing, the method comprising: a) selecting a plurality of directly observed genetic variants for capture by the nucleic acid probe; b) excluding low-confidence variants from the plurality of directly observed genetic variants, thereby creating a filtered plurality of directly observed genetic variants; c) phasing the filtered plurality of directly observed genetic variants; d) identifying the presence or absence of one or more proxy variants for each variant in the filtered plurality of directly observed genetic variants; e) selecting a plurality of candidate regions of genomic DNA comprising the filtered plurality of directly observed genetic variants, wherein each candidate region of genomic DNA comprises about 25 to about 150 bases and comprises at least one variant in the filtered plurality of directly observed genetic variants; f) calculating, for each candidate region of genomic DNA, a quality score that estimates probe capture efficiency and alignment success; g) phasing the genomic DNA. calculating a probe score for each candidate region of genomic DNA by multiplying the quality score by the number of variants captured by the candidate region of genomic DNA, wherein the number of variants captured by the candidate region of genomic DNA is the sum of the number of directly observed variants captured by the candidate region of genomic DNA and the number of corresponding proxy variants in different candidate regions of genomic DNA; h) selecting one or more candidate regions of genomic DNA with the highest probe score for inclusion in a final set of regions of genomic DNA; i) repeating steps g) and h) for unselected candidate regions of genomic DNA for inclusion in a final set of regions of genomic DNA, wherein the number of variants in the unselected candidate regions of genomic DNA is equal to 1) the number of directly observed variants in the unselected candidate regions of genomic DNA excluding all directly observed variants in the selected regions of genomic DNA, and 2) excluding all proxy variants corresponding to the directly observed variants in the selected regions of genomic DNA.and the number of corresponding proxy variants in different candidate regions of genomic DNA, wherein steps g) and h) are repeated until a maximum number of regions of genomic DNA are selected; and j) generating a set of nucleic acid probes complementary to the nucleic acid sequence of each genomic region in the final set of regions of genomic DNA.

[0009] The present disclosure also provides a method of genotyping a DNA sample by sequencing, the method comprising: a) hybridizing a set of nucleic acid probes manufactured as described above to a DNA sample to produce probe-hybridized genomic DNA; b) sequencing the probe-hybridized genomic DNA to generate a plurality of sequencing reads; c) mapping the plurality of sequencing reads to a reference genome; d) calling directly observed variants present in the mapped sequencing reads; and e) imputing unobserved variants from unsequenced regions of the genomic DNA, thereby establishing the genotype of the sample DNA.

[0010] The present disclosure also provides a method of genotyping a DNA sample by sequencing using a set of nucleic acid probes, the method including: a) selecting a plurality of regions of genomic DNA from the DNA sample containing a plurality of directly observed genetic variants; b) identifying a set of nucleic acid probes for hybridization to the plurality of regions of the selected genomic DNA; c) hybridizing the set of nucleic acid probes to the DNA sample to produce genomic DNA hybridized to the probes; d) sequencing the genomic DNA hybridized to the probes to generate a plurality of sequencing reads; e) mapping the plurality of sequencing reads to a reference genome; f) calling the directly observed variants present in the mapped sequencing reads; and g) imputing unobserved variants from unsequenced regions of the genomic DNA, thereby establishing the genotype of the sample DNA.

[0011] This patent or application contains at least one drawing executed in color. Copies of this patent or patent application publication with color drawing(s) will be provided by the Office upon request and payment of the necessary fee. [Brief explanation of the drawings]

[0012] [Figure 1] Imputed Rsq by variant bins are shown for two different observations [one is a global screening array (GSA) and the other is the genotyping by sequencing (GxS) approach described herein] and two in silico versions for comparison [one is labeled "Mock GxS" with all variants in the probes from the observed probe region, and the other is labeled "Mock MEGA" with all variants in the region assayed by the MEGA microarray (containing 1.8M variants)]. [Figure 2] Genotyping assays performed on 223,266 samples, each assessed for coverage at the design site, showed an average call rate of 98.9%, with 99.3% of samples having a call rate of 95% or higher. Call rate is the percentage of sites with actionable genotypes. DETAILED DESCRIPTION OF THE INVENTION

[0013] Provided herein is a general strategy that can be used to efficiently design a set of nucleic acid probes, each probe can target multiple genetic variants, for example, for use in capture-based "genotyping by sequencing" method.This capture-based "genotyping by sequencing" method targets multiple short segments ("target regions", each usually 10 to several hundred base pairs in length) of genome, each of which can contain multiple known genetic variants.Selecting individual targeting variants is inefficient in these experiments.For example, in the worst case scenario, 100,000 short target regions may be required to target 100,000 variants that are each independently selected.In a more desirable scenario, these 100,000 variants are clustered together, and can be captured with a significantly smaller number of probes. For example, a more desirable method would capture only 25,000 short target regions (if each target region contains an average of four variants) or 50,000 short target regions (if each target region contains an average of two variants) and identify a set of 100,000 variants that can be genotyped. Alternatively, a set of probes could identify 100,000 short target regions that capture 200,000-400,000 variants (thus likely to significantly outperform 100,000 target regions selected after independently selecting 100,000 variants).

[0014] The methods described herein identify a small set of genomic regions for sequencing, aiming to approach the comprehensiveness of whole-genome sequencing at significantly reduced cost and effort. These regions are selected to be predicted to perform well in targeted capture experiments. Furthermore, when considered together, these regions contain a set of common genetic variants that accurately summarize variation within the genome for GWAS, ancestry estimation, identification of genetic relatives, estimation of polygenic risk scores, and other applications that currently rely on genotyping arrays.

[0015] The methods described herein provide a sequencing-based alternative to genotyping arrays. They provide better genome coverage across multiple ancestries than standard arrays. Selecting a large number of common variants, such as approximately 1.4M, can enable highly accurate imputation across multiple ancestries. The methods described herein can also cover approximately 4.5M to 5.0M common variants per sample with one or more sequencing reads. The reagents described herein are iteratively refined by applying them to samples of diverse ancestries. Features of the methods described herein include, but are not limited to, generating data in parallel with whole-exome sequencing of each sample, selecting a majority of the 1.4M common variants to enable genome-wide imputation of variation, and targeting peaks in genome-wide association studies where additional variants are known, such as mitochondrial DNA, the Y chromosome, and the MHC. The methods described herein generate high-fidelity genotypes for approximately 1.4M variants per sample. These 1.4M variants have a call rate of approximately 98.9% and an accuracy of approximately 99.7% compared to deep whole-genome sequencing data. These 1.4M variants can be used as an alternative to array genotypes in most applications. The methods described herein are bioinformatically efficient, adding less than approximately 10 hours of CPU time to typical exome processing procedures. Each sample can be processed and handled independently.

[0016] The sequencing-based approach for genotyping described herein is based on the high-throughput DNA capture technology described herein. The DNA capture methodology described herein is highly automated and scaled to process millions of samples per year. High-quality exome data and genotyping can be performed simultaneously, facilitating integration of results. The methods described herein also have the advantage of being able to evolve over time to improve coverage of regions or variants of interest. The methods described herein achieve different sequence coverage and accuracy for high-value variants. The methods described herein maximize tagging and minimize the number of capture targets. The probe sets described herein have been validated and improved by using them on various samples to remove / replace poor targets. Probes were selected and experimentally validated to represent genetic variation across multiple ancestries. The probe sets target approximately 1.5M variant sites per sample, covering approximately 2.6% of the genome.

[0017] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting. The methods described herein provide for the selection and production of a set of nucleic acid probes, each of which efficiently captures a short stretch of DNA that overlaps with the probe and can generate sequencing reads that can be similarly aligned. Furthermore, the methods described herein focus on regions of genomic DNA that harbor genetic variations that allow for successful complementation of nearby unobserved genetic variations (i.e., complemented variants) and / or direct observation of significant mutations.

[0018] The present disclosure provides a method for producing nucleic acid probes for genotyping by sequencing, the method comprising: a) selecting a plurality of directly observed genetic variants for capture by the nucleic acid probe; b) excluding low-confidence variants from the plurality of directly observed genetic variants, thereby creating a filtered plurality of directly observed genetic variants; c) phasing the filtered plurality of directly observed genetic variants; d) identifying the presence or absence of one or more proxy variants for each variant in the filtered plurality of directly observed genetic variants; e) selecting a plurality of candidate regions of genomic DNA comprising the filtered plurality of directly observed genetic variants, wherein each candidate region of genomic DNA comprises about 25 to about 150 bases and comprises at least one variant in the filtered plurality of directly observed genetic variants; f) calculating, for each candidate region of genomic DNA, a quality score that estimates probe capture efficiency and alignment success; g) phasing the genomic DNA. calculating a probe score for each candidate region of genomic DNA by multiplying the quality score by the number of variants captured by the candidate region of genomic DNA, wherein the number of variants captured by the candidate region of genomic DNA is the sum of the number of directly observed variants captured by the candidate region of genomic DNA and the number of corresponding proxy variants in different candidate regions of genomic DNA; h) selecting one or more candidate regions of genomic DNA with the highest probe score for inclusion in a final set of regions of genomic DNA; i) repeating steps g) and h) for unselected candidate regions of genomic DNA for inclusion in a final set of regions of genomic DNA, wherein the number of variants in the unselected candidate regions of genomic DNA is equal to 1) the number of directly observed variants in the unselected candidate regions of genomic DNA excluding all directly observed variants in the selected regions of genomic DNA, and 2) excluding all proxy variants corresponding to the directly observed variants in the selected regions of genomic DNA.and the number of corresponding proxy variants in different candidate regions of genomic DNA, wherein steps g) and h) are repeated until a maximum number of regions of genomic DNA are selected; and j) generating a set of nucleic acid probes complementary to the nucleic acid sequence of each genomic region in the final set of regions of genomic DNA.

[0019] The present disclosure provides a method for designing nucleic acid probes for genotyping by sequencing, comprising: a) selecting a plurality of directly observed genetic variants for capture by the nucleic acid probe; b) eliminating low-confidence variants from the plurality of directly observed genetic variants, thereby creating a filtered plurality of directly observed genetic variants; c) phasing the filtered plurality of directly observed genetic variants; d) identifying the presence or absence of one or more proxy variants for each variant in the filtered plurality of directly observed genetic variants; e) selecting a plurality of candidate regions of genomic DNA containing the filtered plurality of directly observed genetic variants, wherein each candidate region of genomic DNA comprises about 25 to about 150 bases and contains at least one variant in the filtered plurality of directly observed genetic variants; f) calculating a quality score for each candidate region of genomic DNA that estimates probe capture efficiency and alignment success; g) phasing the genomic DNA. calculating a probe score for each candidate region of genomic DNA by multiplying the quality score by the number of variants captured by the candidate region of genomic DNA, wherein the number of variants captured by the candidate region of genomic DNA is the sum of the number of directly observed variants captured by the candidate region of genomic DNA and the number of corresponding proxy variants in different candidate regions of genomic DNA; h) selecting one or more candidate regions of genomic DNA with the highest probe score for inclusion in a final set of regions of genomic DNA; and i) repeating steps g) and h) for unselected candidate regions of genomic DNA for inclusion in the final set of regions of genomic DNA, wherein the number of variants in the unselected candidate regions of genomic DNA is equal to 1) the number of directly observed variants in the unselected candidate regions of genomic DNA excluding all directly observed variants in the selected regions of genomic DNA, and 2) excluding all proxy variants corresponding to the directly observed variants in the selected regions of genomic DNA.and the number of corresponding proxy variants in different candidate regions of genomic DNA, and steps g) and h) are repeated until a maximum number of regions of genomic DNA are selected.

[0020] The subject method involves selecting multiple genetic variants for capture by nucleic acid probes. These selected variants constitute a desired set of "directly observed genetic variants." A "directly observed genetic variant" or "directly observed variant" is a variant present in genomic DNA that is captured by hybridization of at least one probe and subsequently sequenced. The directly observed variant is distinct from the remaining genetic variants, including the complemented variant. While all complemented variants are likely present in the same genomic DNA, they are not captured by hybridization of at least one probe, and therefore, the complemented variant is not subsequently sequenced. The presence of the directly observed variant in the genomic DNA and subsequent sequencing allows for the complementation of the complemented variant.

[0021] The plurality of directly observed genetic variants for capture by nucleic acid probes can include any desired number of known common variants. For example, a set of M known genetic variants can be V1, V2, V3, ... V M The indices m and n vary between 1 and M and are used to designate individual variants. Each variant V m is the known chromosomal position P m and allele A m and each variant V n is the known chromosomal position P n and allele A nIn some embodiments, the plurality of directly observed genetic variants comprises each and every known common variant. In some embodiments, the plurality of directly observed genetic variants is selected from a database of genome-wide association of genetic variants, a database of pharmacogenetic association of genetic variants, a database including genetic variants in entire mitochondrial chromosomes, and / or a database of genetic variants in microarrays, or any combination thereof.

[0022] In some embodiments, the plurality of directly observed genetic variants are selected from one or more databases of genome-wide association of genetic variants. Any of the databases of genome-wide association of genetic variants may be used to identify one or more directly observed genetic variants for inclusion. In some embodiments, the database of genome-wide association of genetic variants is a catalog of known genome-wide association hits (see, for example, the World Wide Web at "ebi.ac.uk / gwas / "). In some embodiments, the source file was "gwas_catalog_v1.0.2-associations_e96_r2019-07-30.tsv." In some embodiments, not all variants in the database of genome-wide association of genetic variants are selected. In some embodiments, variants in the database of genome-wide association of genetic variants are selected based on whether the association of the variant with the trait has a p-value ≦10 -9 In some embodiments, a variant in a genome-wide association database of genetic variants is selected to be among the multiple directly observed genetic variants if it has a p-value >10 -9In some embodiments, the P-value analysis excludes variants present on the Y chromosome and mitochondrial chromosomes. In some embodiments, the number of variants selected from the database(s) of genome-wide association of genetic variants is between about 30,000 and about 45,000. In some embodiments, the number of variants selected from the database(s) of genome-wide association of genetic variants is between about 35,000 and about 40,000. In some embodiments, the number of variants selected from the database(s) of genome-wide association of genetic variants is about 38,000. The number of variants selected from the database(s) of genome-wide association of genetic variants is expected to change over time.

[0023] In some embodiments, the plurality of directly observed genetic variants are selected from one or more databases of genetic variant pharmacogenetic associations. Any of the databases of genetic variant pharmacogenetic associations may be used to identify one or more directly observed genetic variants for inclusion. In some embodiments, the database of genetic variant pharmacogenetic associations is the data published by PharmGKB on pharmacogenetic associations. In some embodiments, all observed single nucleotide polymorphisms (SNPs) in dbSNP that overlap with the gene of pharmacogenetic interest are included. In some embodiments, the number of variants selected from the database(s) of genetic variant pharmacogenetic associations is between about 2,000 and about 10,000. In some embodiments, the number of variants selected from the database(s) of genetic variant pharmacogenetic associations is between about 4,000 and about 6,000. In some embodiments, the number of variants selected from the database(s) of genetic variant pharmacogenetic associations is between about 5,000.

[0024] In some embodiments, the plurality of directly observed genetic variants is selected from one or more databases containing genetic variants in all mitochondrial chromosomes. Any of the databases containing genetic variants in all mitochondrial chromosomes can be used to identify one or more directly observed genetic variants for inclusion. In some embodiments, all mitochondrial chromosomes are tiled end-to-end.

[0025] In some embodiments, the plurality of directly observed genetic variants are selected from one or more databases of genetic variants in one or more microarrays. Any of the databases of genetic variants in microarrays can be used to identify one or more directly observed genetic variants for inclusion. An exemplary database is the microarray variants used by UK Biobank. In some embodiments, the database of genetic variants in microarrays includes genetic variants in the HLA region of chromosome 6, the Y chromosome, two killer cell immunoglobulin-like receptor (KIR) regions on chromosome 19, and pseudoautosomal regions 1 and 2 (Par1 and Par2) on the X chromosome.

[0026] In some embodiments, the database of genetic variants in the microarray comprises genetic variants in the HLA region of chromosome 6. In some embodiments, the database of genetic variants in the microarray comprises genetic variants in the HLA region of chromosome 6, defined as Chr6:28011410-33978119. Of course, equivalent coordinates in alternative human genome assemblies are also included herein.

[0027] In some embodiments, the database of genetic variants in the microarray includes genetic variants on the Y chromosome. In some embodiments, the database of genetic variants in the microarray comprises genetic variants in two KIR regions on chromosome 19. In some embodiments, the database of genetic variants in the microarray comprises genetic variants in two KIR regions on chromosome 19, defined as Chr19:53961144-55367153 and Chr19:110783-760809. Of course, equivalent coordinates in alternative human genome assemblies are also included herein.

[0028] In some embodiments, the database of genetic variants in the microarray includes genetic variants in Par1 and Par2 on the X chromosome. In some embodiments, the database of genetic variants in the microarray includes genetic variants in Par1 and Par2 on the X chromosome defined as ChrX:10425-2774669 and ChrX:155704030-156003450. Of course, equivalent coordinates in alternative human genome assemblies are also included herein. In some embodiments, the number of variants selected from the database(s) of genetic variants in the microarray is between about 700,000 and about 900,000. In some embodiments, the number of variants selected from the database(s) of genetic variants in the microarray is between about 800,000 and about 850,000. In some embodiments, the number of variants selected from the database(s) of genetic variants in the microarray is about 830,000.

[0029] In some embodiments, multi-allelic variants are converted into one or more sets of biallelic variants. The conversion involves two steps: one step converts the abstract variants, and the other converts the individual genotypes. In some embodiments, the multi-allelic genotypes of the original multi-allelic variants are converted into biallelic genotypes for each of the decomposed genetic variants, allowing estimation of linkage disequilibrium coefficients and proxy relationships between the genetic variants. The methods described herein can address multi-allelic variants by decomposing each multi-allelic variant into a set of biallelic variants that are all assigned the same chromosomal location but different alleles. For example, if a particular multi-allelic variant has one reference allele and three alternative alleles, the multi-allelic variant is converted into three sets of biallelic variants (i.e., the reference allele and the first alternative allele, the reference allele and the second alternative allele, and the reference allele and the third alternative allele).

[0030] In some embodiments, to calculate metrics for potential imputation success, the whole genome sequencing dataset from the 1000 Genomes Project (denoted 1KG) was procured. High-coverage (30x) sequencing of 2,504 samples from 26 different populations was released for commercial use in May 2019 by the New York Genome Center (see the world wide web at "internationalgenome.org / data-portal / data-collection / 30x-grch38").

[0031] The subject methods also include filtering out low-confidence variants from the plurality of directly observed genetic variants, thereby creating a filtered plurality of directly observed genetic variants. Filtering out low-confidence variants from the plurality of directly observed genetic variants serves as a quality control to limit selected variants to high-confidence variants. In some embodiments, filtering out low-confidence variants from the plurality of potential directly observed genetic variants retains approximately 15 million variants. Filtering out low-confidence variants from the plurality of directly observed genetic variants can include any one or more of the following:

[0032] In some embodiments, eliminating low-confidence variants from the plurality of directly observed genetic variants comprises eliminating all variants with a minor allele frequency (MAF) below a desired threshold. For example, the allele frequency range is f min From f max The variant in V is f min more than f max The variants may be restricted to those with a minor allele frequency of, for example, f max can be set to 0.50. Furthermore, f min can be 1% (0.01) or 5% (0.05). In some embodiments, the desired threshold is 1% (0.01). In some embodiments, this MAF threshold can be as low as 0.1% (0.001).

[0033] In some embodiments, eliminating low-confidence variants from the plurality of directly observed genetic variants comprises eliminating all variants with more than a desired threshold of missing data, in some embodiments, the desired threshold is 2%.

[0034] In some embodiments, excluding low-confidence variants from a plurality of directly observed genetic variants is achieved by excluding low-confidence variants from a plurality of directly observed genetic variants, where the low-confidence variants are categorized as low-confidence variants with a P-value <10 by a Hardy-Weinberg test in any of the sample populations. -8This includes excluding variants with an association of

[0035] The present methods also include phasing the filtered plurality of potential directly observed genetic variants. In some embodiments, the present methods include phasing all variants observed in a 1000 Genomes sample or another reference panel. Phasing these variants helps methods and algorithms select better-performing "directly observed variants" and "probes." Phasing creates a best estimate of the sequence of variants on each of the two chromosomes per sample. Phasing variants in the 1000 Genomes reference panel (or another panel of reference individuals) improves handling of missing data and estimates of linkage disequilibrium and proxy relationships between variants. In contrast, genotyping only provides information about the counts of specific alleles in both chromosome combinations. For example, a sequence with allele counts 0,1,2,2,1,1 can be phased as two binary sequences representing the two sequences on each chromosome: 0,1,1,1,1,1 and 0,0,1,1,0,0. Phasing of genotype calls can be performed by commercially available software such as SHAPEIT4 (see the world wide web at odelaneau.github.io / shapeit4 / ) using all the usual defaults.

[0036] The subject method also includes identifying the presence or absence of one or more proxy variants for each directly observed variant among the filtered plurality of directly observed genetic variants. Each variant among the filtered plurality of directly observed genetic variants can potentially be a proxy (i.e., a proxy variant) for another variant that is not probed or sequenced (i.e., the proxy variant is imputed into the sample DNA genome based on the presence of the directly observed variant). These proxy relationships are often found between nearby variants in the human genome due to linkage disequilibrium. For example, to describe the proxy relationship between two variants, variant V may be identified as a proxy variant. m and V n Entry R describing the linkage disequilibrium relationship between mn A matrix R can be used that includes: A number of suitable measures of linkage disequilibrium between variants exist and can be used in the methods described herein. In some embodiments, when a directly observed genetic variant and a proxy variant are within 1 MB of each other, and the linkage disequilibrium between the two variants is found to be r 2 A variant in the filtered plurality of directly observed genetic variants has a corresponding proxy variant in another region of genomic DNA if it has a squared correlation above a desired threshold (t) using the measure. The adjustable parameter t represents the minimum amount of linkage disequilibrium required before two variants can be considered proxies for each other. In some embodiments, linkage disequilibrium between two variants is measured by the r of linkage disequilibrium. 2 In some embodiments, linkage disequilibrium between two variants is measured using a correlation squared (t) of at least 0.2 using the r 2 In some embodiments, linkage disequilibrium between two variants is measured using a correlation squared (t) of at least 0.5 using the r 2 In some embodiments, the linkage disequilibrium between two variants is measured using a correlation squared (t) of at least 0.8 using the r2 In some embodiments, the linkage disequilibrium between two variants is measured using a correlation squared (t) of at least 0.9 using the r 2 have a correlation squared (t) of at least 1.0 using the metric. In some embodiments, the proxy variant is present in a different candidate region of genomic DNA than the corresponding directly observed variant. Thus, R mn When the value of is greater than t, two variants V m and V n are proxies for each other.

[0037] Typically, a set of known genetic variants V and their linkage disequilibrium relationships R can be estimated by sequencing or genotyping a small set of individuals. The quality of the regions selected for sequencing improves as the number of individuals in this set increases. Furthermore, the individuals in this set should have diverse ancestries, or at least match the ancestry composition of the individuals being studied using the selected target regions.

[0038] In some embodiments, identifying the presence or absence of one or more proxy variants for each directly observed variant can be performed by linkage disequilibrium software. One such example is emeraLD (see the world wide web at "github.com / statgen / emeraLD"), using normal defaults. Using such software, it is possible to generate a list of pairs of variants that are within 1 Mb of each other and have a squared correlation that exceeds a desired threshold t.

[0039] The subject methods also include selecting multiple candidate regions (i.e., target regions) of genomic DNA for capture by nucleic acid probes. One goal is to select a set of K candidate regions of genomic DNA, T=T1, T2, T3, ...T KThe index k varies between 1 and K and can be used to designate individual candidate regions of genomic DNA. Each candidate region T of genomic DNA is k is the starting position Start(T k ) and the end position End(T k ) and the corresponding probe score Score(T k ) and the probe score represents the expected performance of a candidate region of genomic DNA in a targeted experiment. The candidate region of genomic DNA contains multiple directly observed genetic variants that have been filtered.

[0040] The adjustable parameter L defines the maximum allowable length of each candidate region of genomic DNA, which is the starting position of the candidate region in genomic DNA, Start(T k ) and the end position End(T k) is the distance in bases between the first and second candidate regions. Setting L = 1 results in a strategy similar to pairwise tagging algorithms often used to design standard arrays. In contrast, the methods described herein can use L in the range of 25 to 150. In some embodiments, each candidate region of genomic DNA comprises about 25 to about 150 bases and includes at least one variant in the filtered plurality of directly observed genetic variants. In some embodiments, each candidate region of genomic DNA comprises about 35 to about 140 bases and includes at least one variant in the filtered plurality of directly observed genetic variants. In some embodiments, each candidate region of genomic DNA comprises about 45 to about 130 bases and includes at least one variant in the filtered plurality of directly observed genetic variants. In some embodiments, each candidate region of genomic DNA comprises about 55 to about 125 bases and includes at least one variant in the filtered plurality of directly observed genetic variants. In some embodiments, each candidate region of the genomic DNA comprises about 65 to about 125 bases and comprises at least one variant from the filtered plurality of directly observed genetic variants. In some embodiments, each candidate region of the genomic DNA comprises about 75 to about 125 bases and comprises at least one variant from the filtered plurality of directly observed genetic variants. In some embodiments, each candidate region of the genomic DNA comprises about 85 to about 125 bases and comprises at least one variant from the filtered plurality of directly observed genetic variants. In some embodiments, each candidate region of the genomic DNA comprises about 95 to about 125 bases and comprises at least one variant from the filtered plurality of directly observed genetic variants. In some embodiments, each candidate region of the genomic DNA comprises about 105 to about 125 bases and comprises at least one variant from the filtered plurality of directly observed genetic variants. In some embodiments, each candidate region of the genomic DNA comprises about 120 to about 125 bases.

[0041] In some embodiments, the plurality of candidate regions of genomic DNA comprises about 5 million to about 50 million variants. In some embodiments, the plurality of candidate regions of genomic DNA comprises about 10 million to about 40 million variants. In some embodiments, the plurality of candidate regions of genomic DNA comprises about 20 million to about 30 million variants.

[0042] In some embodiments, the entire plurality of candidate regions of genomic DNA comprises about 1 million to about 100 million base pairs. In some embodiments, the entire plurality of candidate regions of genomic DNA comprises about 5 million to about 75 million base pairs. In some embodiments, the entire plurality of candidate regions of genomic DNA comprises about 10 million to about 50 million base pairs. In some embodiments, the entire plurality of candidate regions of genomic DNA comprises about 20 million to about 40 million base pairs.

[0043] In some embodiments, the plurality of candidate regions of genomic DNA are divided into separate analysis groups. In some embodiments, the plurality of candidate regions of genomic DNA are divided into separate chromosomal analysis groups.

[0044] In some embodiments, the plurality of candidate regions of genomic DNA includes two or more directly observed variants among the filtered plurality of directly observed genetic variants. For example, a candidate region of genomic DNA comprising 120 bases may include four directly observed variants (i.e., V1, V2, V3, and V4). In this scenario, each of the four directly observed variants is present in the region of DNA probed with the nucleic acid probe set. The 120-base candidate region of genomic DNA may start at the position of the first variant (i.e., V1...V2...V3...V4...). The 120-base candidate region of genomic DNA may end at the position of the last variant (i.e., ...V1...V2...V3...V4...). Alternatively, the 120-base candidate region of genomic DNA may start and end at positions other than these variant positions (i.e., ...V1...V2...V3...V4...). There can be many different candidate regions of genomic DNA that contain 120 bases and that contain directly observed variants (i.e., by shifting the start position of the candidate region). Thus, multiple different candidate regions of genomic DNA that contain 120 bases can contain the same directly observed variant(s).

[0045] The subject methods also include calculating a quality score for each candidate region of genomic DNA, which estimates the capture efficiency and alignment success of the probe hybridizing to it. The quality score can be used to determine which probes (and corresponding candidate regions of genomic DNA) should be avoided. As described above, multiple different candidate regions of genomic DNA comprising 120 bases can contain the same directly observed variant(s), and therefore a quality score is calculated for each of these candidate regions of genomic DNA containing the same directly observed variant(s). Additionally, quality scores are calculated for each of other candidate regions of genomic DNA containing different directly observed variant(s). In some embodiments, calculating the quality score includes determining component scores for each of the mappability metric, insertion-deletion metric, and classification metric for the candidate region of genomic DNA. The quality score combines these three pieces of information to allow for the reverse mapping of probes and subsequent sequencing reads that perform well in capturing the appropriate strand of DNA, avoiding regions containing insertion-deletion polymorphisms or mutations, and aiming to preferentially select regions that perform well according to the expected performance of probe hybridization to DNA, which can be estimated as a function of sequence composition and uniqueness. The quality score for each candidate region of genomic DNA is the product of the multiplication of each of the component scores for that candidate region of genomic DNA. The final result is a quality score between 0 and 1 that correlates with the probability of probe success. If any of the component scores is zero, the overall quality score is also zero.

[0046] In some embodiments, the mappability metric (or multiread mappability metric) is the probability that a randomly selected read of length k in a given region is uniquely mappable. In some embodiments, the mappability metric is the UMAP metric. In some embodiments, the component scores of the mappability metric are the multiread mappability metric (UmapMRM for position i)i In some embodiments, the component scores of the mappability metric are expressed as exp(10×UmapMRM i -9), where UmapMRM i is the multi-read mappability metric for variant position i within a candidate region of genomic DNA. In some embodiments, UMAP mapping metrics, particularly 100 bp multi-read mappability metrics, have been pre-calculated genome-wide and compiled in downloadable tables (see the world wide web at "bismap.hoffmanlab.org / ").

[0047] In some embodiments, the insertion-deletion metric is a measure of the presence or absence of an insertion or deletion of bases (e.g., an insertion-deletion polymorphism or mutation) within a candidate region of genomic DNA. The insertion-deletion is included as if position i were linked to the insertion-deletion mutation, and then this position is downweighted. In some embodiments, the component score of an insertion-deletion mutation is calculated as exp(SV score i In some embodiments, if variant position i is not linked to an insertion-deletion mutation or is linked to an insertion-deletion mutation of less than 5 bases, then SV score i is 2. In some embodiments, if variant position i is linked to an insertion-deletion mutation of 5 to 10 bases (e.g., a medium-sized insertion-deletion variant), the SV score i is 1. In some embodiments, if variant position i is linked to an insertion-deletion mutation of more than 10 bases (e.g., a large insertion-deletion), the SV score i is 0. In some embodiments, if the variant position is not near an insertion-deletion variant, the SV score i is 2, and if the variant position is near an insertion-deletion variant of ≥ 5 and < 10 bases, the SV score is i is 1, and if the variant position is near an insertion-deletion variant of ≥ 10 bases, the SV score is iis 0. An adjustable parameter can define the maximum length of an insertion-deletion polymorphism included in the candidate region of genomic DNA. This adjustable parameter can depend on the tolerance for mismatches between the probes used for targeting and the sequences present in each sample being investigated.

[0048] In some embodiments, the classification metric for a candidate region of genomic DNA includes a first category (e.g., the poorest performing category), a second category (e.g., a poorly performing category), a third category (e.g., a poorly performing category), and a fourth category (e.g., a good performing category). The order from best performing to worst performing is the fourth category, the third category, the second category, and the first category. In some embodiments, the first component score of the classification metric is exp(Region_score i ), whereby variant position i in the first category is scored as 0, variant position i in the second category is scored as 1, variant position i in the third category is scored as 1.6, and variant position i in the fourth category is scored as 2. In some embodiments, the second component score, which is the minimum absolute distance score of the classification metric, is

[0049]

number

[0050] where dist2category1 i is the minimum absolute distance from variant position i to the region in the first category. In some embodiments, the third component score of the classification metric is

[0051]

number

[0052] where dist2category2i is the minimum absolute distance from variant position i to a region in the second category. These two component scores downweight probes that are not in category 1 or category 2 (i.e., bad or worst-case regions) but are very close, so that reads made from those probes may have poor alignments.

[0053] In some embodiments, the trait used to categorize a particular candidate region of genomic DNA may be the %GC content of the corresponding complementary probe / primer. For example, the %GC content of the probe / primer is preferably about 40% to about 55%. Thus, in some embodiments, a first category may have corresponding probes / primers with a %GC content of less than about 40%, a second category may have corresponding probes / primers with a %GC content of greater than 55%, a third category may have corresponding probes / primers with a %GC content of about 50% to about 55%, and a fourth category may have corresponding probes / primers with a %GC content of about 40% to about 55%. Additional traits that can be used to categorize a particular candidate region of genomic DNA include, but are not limited to, primer / probe melting temperature, primer / probe annealing temperature, the presence or absence of a GC clamp, 3' end stability, etc. Each of these traits may be divided into four categories based on the user's desired preferences.

[0054] The overall quality score is the product of the multiplication of the five component scores. In some embodiments, the quality score for each candidate region of genomic DNA is calculated by multiplying the maximum score (exp(5)×1.2 2 ; or approximately 213.7149) to scale it from 0 to 1, thereby creating a quality score for each candidate region of genomic DNA.

[0055] With respect to the overall quality score, decisions made about which probes to select for a particular candidate region of genomic DNA can be relative. Thus, regional characteristics (such as GC content) that lower the scores of many nearby probes do not necessarily eliminate that region from consideration. Rather, our method attempts to select the best probes available in such a region. Additionally, the quality score can also include a metric that prioritizes probes that are evenly distributed across the genome.

[0056] The method also includes calculating a probe score for each candidate region of genomic DNA. In some embodiments, the probe score is calculated by multiplying the quality score by the number of variants captured by the candidate region of genomic DNA. For example, for each candidate region of genomic DNA, T k may overlap with the set of genetic variants, which is called OverlapSet(T k ), and Start(T k ) and End(T k ) for each candidate region of genomic DNA, including all genetic variants located between T k In addition to the variants it directly overlaps, it also includes the OverlapSet(T k ) and also capture variants that have proxies in the region T k This can be called a proxy set of ProxySet(T k ), and OverlapSet(T k ), as well as all variants in R mn > t so that OverlapSet(T k ) includes all other variants m for which there is a corresponding variant n in . Thus, in some embodiments, the number of variants captured by a candidate region of genomic DNA is the sum of the number of directly observed variants (i.e., in the candidate region hybridized to the probe) captured by the candidate region of genomic DNA and the number of corresponding proxy variants in different candidate regions of genomic DNA.

[0057] For example, if a particular candidate region of genomic DNA contains three directly observed variants (i.e., V1, V2, and V3), V1 has two corresponding proxy variants, PV1 and PV2. a and PV b in different candidate regions of genomic DNA, and V2 is the four corresponding proxy variant PVs. c , PV d , PV e , and P.V. f in different candidate regions of genomic DNA, and V3 has five corresponding proxy variants PV g , PV h , PV i , PV j , and P.V. k in different candidate regions of genomic DNA, the number of directly observed variants captured by the candidate regions of genomic DNA is 3 (i.e., V1, V2, and V3), and the number of corresponding proxy variants in different candidate regions of genomic DNA is 11 (i.e., PV a , PV b , PV c , PV d , PV e , PV f , PV g , PV h , PV i , PV j , and P.V. k ) Therefore, the number of directly observed variants captured by a candidate region of genomic DNA plus the number of corresponding proxy variants in a different candidate region of genomic DNA is 14. Therefore, the probe score for this particular candidate region of genomic DNA is the product of the quality score multiplied by 14.

[0058] The subject methods also include selecting one or more candidate regions of genomic DNA with the highest probe scores for inclusion in the final set of regions of genomic DNA. In some embodiments, the single candidate region of genomic DNA with the highest probe score is selected for inclusion in the final set of regions of genomic DNA. In some embodiments, two or more candidate regions of genomic DNA with the highest probe scores are selected for inclusion in the final set of regions of genomic DNA. In some embodiments, when there are multiple candidate regions of genomic DNA with the highest probe scores, candidate regions of genomic DNA that are more evenly spaced across the genome are selected.

[0059] In selecting a set of candidate regions of genomic DNA to experimentally measure, one goal is to minimize the number of regions in T that yield an overall probe score, Score(T k ) and maximize the overall quality of these regions summarized by the ProxySet (T k ) to maximize the number of variants captured in the union. When there are multiple sets of candidate genomic DNA regions that perform similarly, sets of evenly spaced candidate genomic DNA regions can be prioritized, as these sets are likely to actually outperform alternatives.

[0060] As described herein, one step in the methods described herein is the identification of a set of candidate regions of genomic DNA to be evaluated. Because the human genome is approximately 3 billion base pairs long, there are 3×10 potential candidate regions of genomic DNA of length L. 9There are likely to be around 50 million variants (if L is small compared to the genome size). The number of candidate variants potentially selected is much smaller, typically around 50-50 million variants (depending on the allele frequency range of the variants). The list of candidate regions in genomic DNA is seeded with a recommended candidate region in genomic DNA for each variant. This recommended candidate region in genomic DNA includes this variant and all variants within L base pairs to its right. Among all possible candidate regions in genomic DNA that meet this criteria, the one with the highest probe score, Score(T k The focus is on suggested candidate regions of genomic DNA with the same base pair as the target region. Performance can be improved by also considering regions L base pairs to the right, but containing only a subset of variants with higher region probe scores. For example, variant V m as well as three additional variants V m+1 , V m+2 , and V m+3 are all within L base pairs to its right. Without loss of generality, we can sort the three variants from left to right according to their coordinates: V m , V m+1 , V m+2 , and V m+3 and identify the candidate region with the highest possible score. m , V m+1 , and V m+2 Only or V m and V m+1 The highest scoring candidate regions containing only V m , V m+1 , V m+2 , and V m+3Only if the probe score of the highest-scoring region containing the region is higher than the probe score of the region containing the highest-scoring region, are they added to the list of potential candidate regions in the genomic DNA. If these additional regions have low region probe scores, they will never be selected and can be safely ignored, since the list of variants for which they serve as proxies will always be smaller than or equal to the list of regions that the higher-scoring regions can proxy. This optional step reduces the number of candidate regions in the genomic DNA that need to be considered in each iteration from billions to millions, resulting in significant savings in computational time.

[0061] In some embodiments, an additional adjustable parameter can be used to define the maximum number of variants allowed per candidate region of genomic DNA. In some embodiments, if a candidate region of genomic DNA contains more directly observed variants than a desired threshold, the candidate region of genomic DNA is removed from the final set of regions of genomic DNA. In some embodiments, the desired threshold is five directly observed variants.

[0062] The subject methods also include repeating steps g) (i.e., calculating a probe score for each candidate region of the genomic DNA) and h) (i.e., selecting one or more candidate regions of the genomic DNA with the highest probe scores for inclusion in the final set of regions of the genomic DNA) for unselected candidate regions of the genomic DNA for inclusion in the final set of regions of the genomic DNA. Thus, to identify a set of candidate regions of the genomic DNA, the methods described herein proceed through an iterative series of steps. In each iteration, one or more candidate regions of the genomic DNA are selected for inclusion in the final set of candidate regions of the genomic DNA, and the scores of other candidate regions of the genomic DNA are updated. The selection of candidate regions of the genomic DNA for inclusion in the final set of candidate regions of the genomic DNA continues until a maximum number of candidate regions of the genomic DNA have been selected, or until all variants of interest are located within or have proxies within the selected candidate regions of the genomic DNA.

[0063] For example, after the initial selection of a single or multiple candidate regions of genomic DNA described in the previous step, the remaining unselected candidate regions of genomic DNA become available for recalculation of probe scores and selection for inclusion in the final set of regions of genomic DNA. In such repetition of steps, the number of variants in a particular unselected candidate region of genomic DNA is the sum of 1) the number of directly observed variants in the unselected candidate region of genomic DNA, excluding all directly observed variants in the selected candidate region of genomic DNA, and 2) the number of corresponding proxy variants in a different candidate region of genomic DNA, excluding all proxy variants corresponding to the directly observed variants in the selected candidate region of genomic DNA.

[0064] For example, assume that a selected candidate region of genomic DNA (i.e., candidate region 1 from step h) contains two directly observed variants (i.e., V1 and V2). Also, assume that V1 has two corresponding proxy variants PV a and PV b in different candidate regions of genomic DNA, and V2 is the two corresponding proxy variants PV c and PV d Suppose that candidate region 2 under consideration for selection has two directly observed variants (i.e., V2 and V3), where V2 has two corresponding proxy variants PV c and PV d in different candidate regions of genomic DNA, and V3 is the two corresponding proxy variant PVs. e and PV fSuppose that candidate region 2 has V2 and V3 in a different candidate region of genomic DNA. When candidate region 2 is considered for selection, the number of directly observed variants in the unselected candidate region 2 excludes all directly observed variants in the candidate region of the selected genomic DNA (i.e., V2 from candidate region 1), and the number of corresponding proxy variants in the different candidate region of genomic DNA excludes all proxy variants corresponding to directly observed variants in the candidate region of the selected genomic DNA (i.e., proxy variant PV associated with V2 from candidate region 1). c and PV d ) are excluded. Therefore, in the scenario described herein, candidate region 2 contains two directly observed variants (i.e., V2 and V3), but only one of them (i.e., V3) is counted toward the number of directly observed variants for determining the probe score. Furthermore, candidate region 2 contains four proxy variants (i.e., PV c , PV d , PV e , and P.V. f ), but only two of them (i.e., PV e and PV f ) are counted towards the number of corresponding proxy variants to determine the probe score. Thus, in the current scenario, rather than the probe score for candidate region 2 being the multiplication product of the quality scores for candidate regions 2 and 6 (i.e., the sum of the two directly observed variants and the four corresponding proxy variants), the probe score for candidate region 2 is the multiplication product of the quality scores for candidate regions 2 and 3 (i.e., the sum of the single directly observed variant and the two corresponding proxy variants that are not already present in any of the candidate regions of the selected DNA).

[0065] In some embodiments, after steps g) (i.e., calculating a probe score for each candidate region of the genomic DNA) and h) (i.e., selecting one or more candidate regions of the genomic DNA with the highest probe scores for inclusion in the final set of regions of the genomic DNA) are repeated, the probe scores of the remaining unselected candidate regions of the genomic DNA are updated.

[0066] In some embodiments, updating involves selecting candidate genomic DNA regions for inclusion in the final set of regions of genomic DNA, and then recalculating probe scores for all remaining unselected candidate genomic DNA regions that contain proxies for directly observed variants that were present in the previously selected candidate genomic DNA regions. In some embodiments, updating involves eliminating all unselected candidate genomic DNA regions that contain only directly observed variants and / or corresponding proxy variants that were previously selected for inclusion in the final set of regions of genomic DNA in a previous round of selection. In some embodiments, updating involves both of the above updates.

[0067] In some embodiments, steps g) and h) are repeated until a maximum number of regions of genomic DNA are selected, hi some embodiments, steps g) and h) are repeated until all directly observed variants and proxy variants are included in the final set of regions of genomic DNA.

[0068] All potential candidate regions of genomic DNA are repeated in each iteration. k Probe score Score(T k ) and the proxy set ProxySet(T k) and the number of variants in the region. One goal is to identify and select the genomic DNA candidate region with the highest incremental value. If there is a tie, it is broken by selecting the genomic DNA candidate region whose distance between the tied genomic DNA candidate region with the largest product and all of the selected genomic DNA candidate regions and the tie is furthest from the selected genomic DNA candidate region. This tie-breaking strategy promotes even spacing of the selected genomic DNA candidate regions throughout the genome and improves the performance of the methodology when the resulting genomic DNA candidate regions and data analysis are combined with modern haplotyping and imputation methodologies.

[0069] After selecting the candidate region of genomic DNA with the highest increment value and breaking ties if necessary, information about the remaining candidate regions of genomic DNA may be updated. For example, two optional updates may be considered. First, the number of variants in the proxy set for each candidate region of genomic DNA that are not in the proxy set for the previously selected candidate region of genomic DNA may be cached. While this caching is not required, it significantly improves computational efficiency. When caching is enabled, the number of variants in a particular candidate region of genomic DNA, T k After selecting the proxy set, the proxy set is ProxySet(T k ) and updates the cached count of the number of variants in the proxy set that are not within the candidate region of the selected genomic DNA, so that some of the variants in the proxy set are within the candidate region T of the selected genomic DNA. k Second, if the probe score for each candidate region of genomic DNA depends on the probe scores of other selected candidate regions of genomic DNA (e.g., because the targeting technology being used does not allow overlapping regions or because sequence complementarity between the candidate regions of genomic DNA being targeted must be considered), the probe scores for the other candidate regions of genomic DNA will be reflected in the probe scores for the candidate region T of genomic DNA. k may be updated to reflect that a selection has been made.

[0070] Before starting the next iteration, all genomic DNA candidate regions whose proxy set is empty or that are completely contained in the union of the proxy sets of the currently selected genomic DNA candidate regions can be removed from the list of genomic DNA candidate regions to be evaluated. If caching is implemented, these regions have a cache score of zero. These regions will never be selected because they do not improve the design and can be safely removed from the list of genomic DNA candidate regions to be evaluated to improve computational efficiency and speed up future iterations. Furthermore, genomic DNA candidate regions with a cache score of 1 (i.e., capturing only a single incremental variant) can be safely reserved for evaluation in a final custom iteration if the captured variant is not captured by any other genomic DNA candidate regions. This methodology can proceed iteratively, selecting one genomic DNA candidate region at a time, until all variants are included in the proxy set of one genomic DNA candidate region selected for targeting, or until the maximum number of genomic DNA candidate regions has been targeted.

[0071] The methods described herein can be incorporated into algorithms. Additional information can also be used to improve the computational efficiency of the algorithm. For example, a challenging aspect of such an algorithm can be the storage of the matrix R. When the number of variants considered, M, is large, the number of entries in this matrix, proportional to M x M, can become very large and exceed the random access memory (RAM) capacity of most modern computers. In such situations, a sparse representation of the matrix can be used, with only entries with values ​​exceeding a user-defined threshold t that establishes a proxy relationship being loaded into RAM. In typical human data, large linkage disequilibrium coefficients are limited to a small number of variant pairs, and this sparse representation of the matrix can be easily stored in memory and used for the required calculations.

[0072] Furthermore, while the algorithm may be efficient enough to apply directly to the entire genome, selecting candidate regions of genomic DNA for targeting can also be considered, particularly in situations where it does not affect the probe scores of other distant candidate regions of the genomic DNA under consideration, and several efficiencies may be improved. One of these efficiencies is to partition the genome into a series of regions where candidate regions of genomic DNA can be selected independently. In the simplest case, these regions can be individual chromosomes. In a more refined case, the genome can be partitioned into a series of non-overlapping regions such that R mn is guaranteed to be <t> when indexing variants within regions where m and n are different. This partitioning can be done using standard algorithms to identify connected components within a graph. The partitioning improves computational efficiency and allows the algorithm to consider pairs, triples, or other small tuples of candidate regions of genomic DNA for each iteration rather than just one candidate region of genomic DNA for each iteration.

[0073] Iterative algorithms can provide very high-quality solutions that take into account known linkage disequilibrium relationships, prioritize clustered variants that can be targeted together because they fall within a contiguous window of L base pairs or less, tolerate probe scores for candidate genomic DNA regions, and distribute probes evenly across the genome, all in a computationally efficient manner. When the number of candidate genomic DNA regions is modest (or an algorithm is used that divides the genome into blocks that can be considered independently), it is possible to exhaustively enumerate and evaluate all possible combinations of candidate genomic DNA regions. In this case, a global scoring scheme can be used to select the optimal combination of candidate genomic DNA regions from among all the enumerated possibilities. To do this, the global scoring scheme can summarize the number of variants with proxies within the candidate genomic DNA regions, the overall probe score for the candidate genomic DNA regions, and the even spacing of the candidate genomic DNA regions. Given a set T of candidate genomic DNA regions, many suitable scoring schemes can be devised. Each variant of interest can be assigned the probe score of the genomic DNA candidate region that scores highest among the selected genomic DNA candidate regions containing variants in the proxy set. Variants not included in any proxy set can be assigned a score of zero. The overall global score for each configuration can then be a weighted sum of the scores assigned to each of these variants (summed over all variants), a measure of evenness of spacing of the genomic DNA candidate region, such as the kurtosis of the distribution of distances between consecutive selected probes, and a penalty to favor configurations with a small number of targets. This global scoring scheme can also be used in conjunction with pseudo-annealing or another Monte Carlo algorithm to refine the iterative solution recommended by the algorithm. This refinement can also be possible in situations where the set of all possible combinations of genomic DNA candidate regions is too large to enumerate.Like other Monte Carlo schemes, pseudo-annealing requires a proposal scheme to explore solutions in the neighborhood of the current solution and recommend new solutions in the neighborhood of the current solution (e.g., by adding, removing, or substituting candidate regions of genomic DNA in the currently selected set), a scheme to accept or reject the proposed updates probabilistically (e.g., by always accepting solutions that improve the global score and occasionally accepting solutions that decrease the global score to avoid getting stuck in a local minimum), and a scheme to manage the stochastic component of the process so that it becomes progressively more stringent and to determine when convergence has been achieved.

[0074] The subject methods also optionally include generating a set of nucleic acid probes. Each individual probe in the set of nucleic acid probes is complementary to a nucleic acid sequence of a genomic region in the final set of selected genomic DNA regions. Thus, the entire set of nucleic acid probes is complementary to the entire nucleotide sequence of the final set of selected genomic DNA regions. In some embodiments, the set of nucleic acid probes comprises between about 200,000 and about 700,000 probes. In some embodiments, the set of nucleic acid probes comprises between about 200,000 and about 600,000 probes. In some embodiments, the set of nucleic acid probes comprises between about 200,000 and about 500,000 probes. In some embodiments, the set of nucleic acid probes comprises between about 200,000 and about 400,000 probes. In some embodiments, the set of nucleic acid probes comprises between about 500,000 and about 700,000 probes. In some embodiments, the set of nucleic acid probes comprises between about 600,000 and about 650,000 probes. In some embodiments, each individual probe in the set of nucleic acid probes contains about 25 to about 150 bases and is hybridizable to a specific candidate region of genomic DNA containing at least one directly observed variant. In some embodiments, each individual probe in the set of nucleic acid probes contains about 120 to about 125 bases. In some embodiments, one or more individual probes in the set of nucleic acid probes contain the same number of bases as the corresponding candidate region of genomic DNA to which it is designed to hybridize. In some embodiments, one or more individual probes in the set of nucleic acid probes contain a greater number of bases than the corresponding candidate region of genomic DNA to which it is designed to hybridize.

[0075] The present disclosure also provides a method of genotyping a DNA sample by sequencing, the method comprising: a) hybridizing a set of nucleic acid probes manufactured as described herein to the DNA sample to produce genomic DNA hybridized to the probes; b) sequencing the genomic DNA hybridized to the probes to generate a plurality of sequencing reads; c) mapping the plurality of sequencing reads to a reference genome; d) calling directly observed variants present in the mapped sequencing reads; and e) imputing unobserved variants from unsequenced regions of the genomic DNA, thereby establishing the genotype of the sample DNA.

[0076] DNA sample can be any DNA sample that is the DNA source for genotyping.In some embodiments, DNA sample is obtained from the subject with disease or condition.In some embodiments, DNA sample is obtained from the tumor of the subject.

[0077] The method includes hybridizing a set of nucleic acid probes prepared as described herein to a DNA sample to produce probe-hybridized genomic DNA. The set of nucleic acid probes is contacted with the DNA sample under typical hybridization conditions. In some embodiments, if the average probe provides coverage of X, probes with coverage <0.33X can be removed. Thus, for example, all probes that provide less than 8X coverage of directly observed variants among multiple sequencing reads (if the average probe has 24X coverage) are removed from the set of nucleic acid probes. In some embodiments, all probes that result in inefficient capture of sample DNA are removed from the set of nucleic acid probes. In some embodiments, probes that provide low average coverage but target high-value variants (for mapping to known functional regions of the genome or serving as proxies for many other variants) can be supplemented with additional copies in the capture reagent rather than being truncated. This supplementation can help improve the coverage they provide and facilitate accurate genotyping.

[0078] The subject methods also include sequencing the genomic DNA hybridized to the probes to generate a plurality of sequencing reads. In some embodiments, the plurality of sequencing reads comprises about 30 million sequencing reads. In some embodiments, the plurality of sequencing reads comprises about 25 million sequencing reads. In some embodiments, the plurality of sequencing reads comprises about 20 million sequencing reads. In some embodiments, the plurality of sequencing reads comprises about 15 million sequencing reads. In some embodiments, the plurality of sequencing reads comprises about 10 million sequencing reads. In some embodiments, the plurality of sequencing reads comprises about 5 million sequencing reads. In some embodiments, the plurality of sequencing reads comprises about 1 million sequencing reads.

[0079] The subject methods also include mapping the multiple sequencing reads to a reference genome. The method also includes calling directly observed variants present in the mapped sequencing reads. In some embodiments, low-confidence called variants resulting from low-coverage reads are eliminated to create a final set of called directly observed variants. In some embodiments, low-confidence called variants resulting from reads with less than 8X coverage are eliminated. In some embodiments, eliminating low-confidence called variants includes imputing the same called directly observed variants from a reference panel of variants.

[0080] In some embodiments, the subject methods further comprise phasing the called directly observed variants into a set of known haplotypes. Examples of phasing can be found, for example, in U.S. Patent Application Publication No. 2019 / 0205502.

[0081] In some embodiments, the software GLIMPSE (see the world wide web at "odelaneau.github.io / GLIMPSE / "), or software providing the same functionality, can be used to return refined variant calls after including information from nearby variants. GLIMPSE can significantly reduce the uncertainty of variant calls from low-coverage reads, given nearby variant calls for each sample. The second step of GLIMPSE is to take these refined variant calls and phase the genotype calls into variant calls per chromosome. GLIMPSE can be run using default parameters.

[0082] In some embodiments, the percentage of called variants with greater than 10X coverage is determined. In such embodiments, if the percentage of called variants with greater than 10X coverage is less than about 95%, the set of nucleic acid probes is rehybridized to the DNA sample. This embodiment serves as an internal control for the hybridization and sequencing steps described herein.

[0083] In some embodiments, when a called directly observed variant is near or within a region of genomic DNA that is hybridizable to a probe that is excluded from the set of nucleic acid probes, such directly observed variant is removed from the final set of called directly observed variants.

[0084] The subject methods also include imputing unobserved variants from unsequenced regions of genomic DNA, thereby establishing the genotype of the sample DNA. In some embodiments, unobserved variants are imputed from a reference panel of variants based on the presence of called directly observed variants in the DNA sample.

[0085] In some embodiments, the software Minimac3 (see the world wide web at "genome.sph.umich.edu / wiki / Minimac3") can be used for variant imputation (for unobserved and unsequenced variants) from the variant calls of each haplotype. Minimac3 can be run using default parameters.

[0086] The present disclosure also provides a method for genotyping a DNA sample by sequencing using a set of nucleic acid probes, the method comprising: a) selecting multiple regions of genomic DNA from the DNA sample containing multiple directly observed genetic variants; b) identifying a set of nucleic acid probes for hybridization to the selected multiple regions of genomic DNA; c) hybridizing the set of nucleic acid probes to the DNA sample to produce probe-hybridized genomic DNA; d) sequencing the probe-hybridized genomic DNA to generate multiple sequencing reads; e) mapping the multiple sequencing reads to a reference genome; f) calling the directly observed variants present in the mapped sequencing reads; and g) imputing unobserved variants from unsequenced regions of genomic DNA, thereby establishing the genotype of the sample DNA. Steps a) through g) can be performed according to the disclosure herein.

[0087] The present disclosure also provides systems and computer-readable media for performing the methods described herein. In some embodiments, a computer program product is provided, comprising a computer-readable medium containing encoded instructions for performing any of the methods described herein. In some embodiments, the computer program product can cause a computer having a processor to perform any of the methods described herein. In some embodiments, the computer program product is encoded such that the program, when implemented by a suitable computer or system, can receive all parameters necessary to perform any of the methods described herein. In some embodiments, a computer system for performing any of the methods described herein is provided, comprising a processor and a memory coupled to the processor, the memory encoding one or more computer programs that cause the processor to perform any of the methods described herein.

[0088] The computer software product can be written using any suitable programming language known in the art. The system components can include any suitable hardware known in the art. Suitable programming languages ​​and suitable hardware system components are described in U.S. Pat. No. 7,197,400 (see, e.g., columns 8-9); U.S. Pat. No. 6,691,042 (see, e.g., columns 12-25); U.S. Pat. No. 8,245,517 (see, e.g., columns 16-17); U.S. Pat. No. 7,272,584 (see, e.g., column 4, line 26 to column 5, line 18); U.S. Pat. No. 8,245,517 (see, e.g., column 16-17); U.S. Pat. No. 7,272,584 (see, e.g., column 4, line 26 to column 5, line 18); U.S. Pat. No. 03,987 (see, for example, columns 19-20); U.S. Pat. No. 7,386,523 (see, for example, column 2, line 26 to column 3, line 3; also see column 8, line 21 to column 9, line 52); U.S. Pat. No. 7,353,116 (see, for example, column 5, line 50 to column 8, line 5), and U.S. Pat. No. 5,985,352 (see, for example, column 31, line 37 to column 32, line 21).

[0089] In some embodiments, a computer system capable of performing the computer-implemented methods herein comprises a processor, a fixed storage medium (i.e., a hard drive), system memory (e.g., RAM and / or ROM), a keyboard, a display (e.g., a monitor), a data input device (e.g., a device capable of providing raw or converted microarray data to the system), and optionally a drive capable of reading and / or writing computer-readable media (i.e., a removable storage device, e.g., a CD or DVD drive). The system also optionally comprises a network input / output device and a device allowing connection to the Internet.

[0090] In some embodiments, computer-readable instructions (e.g., a computer software product) that enable the system to perform any of the methods described herein (i.e., software for performing any of the method steps described herein) are encoded on a non-transitory storage medium, enabling the system to display the results to a user, or provide the results to a second set of computer-readable instructions (i.e., a second program), or transmit the results to a data structure residing on the non-transitory storage medium, or to another networked computer, or to a remote location via the Internet.

[0091] In order that the subject matter disclosed herein may be more efficiently understood, the following examples are provided, it being understood that these examples are for illustrative purposes only and are not to be construed as limiting the claimed subject matter in any manner. [Example]

[0092] Example 1: Pilot Study After selecting directly observed variants, selecting candidate regions of genomic DNA containing the selected directly observed variants, and selecting probe sets as described herein, a pilot study was performed.

[0093] Forty-eight samples were selected from the 1KG sample set, and these DNA samples were accessed from Coriell (see the World Wide Web at "coriell.org / 1 / NHGRI / Collections / 1000-Genomes-Collections / 1000-Genomes-Project"). In this example, the 48 samples were treated as if they were completely new and processed with the sequencing genotyping probe set described herein. The sequencing genotyping results of the 48 samples were compared with control results obtained from whole genome sequencing at 30X coverage (after filtering). The reference panel was considered to be the 1KG WGS data excluding the 48 samples.

[0094] The pilot set of samples was selected to be diverse. One sample was eliminated because it did not have enough DNA to sequence, leaving 47 samples for testing. The samples are summarized in Table 1.

[0095] [Table 1]

[0096] The first objective was to determine how well the probes actually performed (i.e., whether the probe set captured sequences specific to the desired location in the genome). Two reasons were considered for eliminating certain probes from the initial probe set: 1) the variant coverage was too low, such that some DNA samples did not produce a signal, and 2) many reads were shown not to map readily to the genome at the point captured by that probe. The overall goal was to eliminate probes that resulted in inefficient capture and to eliminate probes that did not provide sufficient signal for the desired variants. Many probes fell into both categories. As a result, approximately 14,000 probes were identified that achieved too low coverage.

[0097] Computational experiments were performed that showed that the excluded probes did not make a significant difference to the overall imputation performance, and this data was observed by filtering the WGS experiments to represent what might be observed.

[0098] Another objective was to determine whether information extracted from sequencing reads could supplement directly observed variants and enable the imputation of other variants. To assess the accuracy of imputation, we performed two processes: 1) we excluded variants close to or within the excluded probes from the called variants, and 2) we processed the remaining called variants to return imputed variants (for all estimated 15 million variants).

[0099] Data Preparation Methods – Variant Calling for Imputation A new reference set of haplotypes was used to perform imputation on the pilot samples. The reference was a 1KG WGS dataset with the pilot samples removed. This new reference data was used twice: 1) by the program GLIMPSE to improve variant calling and phasing, and 2) by the program Minimac3 for variant imputation. The imputed variant calls were then compared to the variant calls observed directly from whole-genome sequencing.

[0100] Evaluation of completion quality To assess imputation quality, we evaluated the squared correlation between the directly observed genotypes and the imputed genotypes. This metric is commonly referred to as "imputation Rsq" or "r 2This is called the "r-measure" or "r-squared," and is the square of the correlation coefficient between true genotypes and their experimentally derived counterparts, as estimated from imputation. When r2 is 1.0, the two are identical. When it is close to 0.0, the experimentally derived counterparts are no better than blind estimates. Specifically, from the whole-genome sequencing data, we created genotype vectors of directly observed genotypes, encoded as 0 if the genotype is for two reference alleles, 1 if the genotype is for one reference and one alternative allele, and 2 if the genotype is for two reference alleles. For the imputed genotype vector, this was different, as each of the three states has a probability. For example, there could be an 80% chance of 0, a 20% chance of 1, and a 0% chance of 2. For the imputed genotype vector, the probability was 0.8 * 0+0.2 * 1+0 * 2 returned a genotype prediction of 0.2.

[0101] Pearson's correlation coefficient was used on the two vectors. Note that there were only 47 samples per genotype. To improve the measurement across variants, variants were pooled together by allele frequency (so that all had the same expected genotype) and vector correlation between samples and variants was performed. This imputation Rsq process followed standard procedures.

[0102] Figure 1 shows the imputation Rsq of difference frequency bins by imputation from different observational data. The highest correlation (and best imputation) occurred when whole-genome sequencing was filtered to observe only variants within the selected probe region. The line formed in this way represented the best performance desired. The blue line represents the global screening array assayed directly with these samples (performed in-house under standard protocols). Imputation from the pilot study was expected to be at least as good as the global screening array. The green line represents the imputation quality of the directly observed sequencing-based genotyping design after processing as described herein. The sequencing-based genotyping design performed significantly better than the global screening array and was close to the best performance desired, given the probes selected. This pilot study demonstrates that sequencing-based genotyping designs can outperform global screening arrays at a reasonable cost. The pilot study was not simply a simulation study; it directly compared the performance of the two assays from DNA samples to imputation comparisons. Finally, we compared our genotyping-by-sequencing design to a very large array called the MEGA array (Multi-Ethnic Genotyping Array), which has three times the number of variants as the global screening array. When we simulated the array by fully observing all variants that the array assays from a whole-genome sequencing version of the pilot data, our genotyping-by-sequencing design performed similarly to the best possible performance of the MEGA array. In fact, the MEGA array performed worse. Our genotyping-by-sequencing design performed similarly to the MEGA array at a cost comparable to the global screening array (one-third that of the MEGA array). Thus, our genotyping-by-sequencing design provided a very cost-effective strategy for assaying genetic information and performed well to produce high-quality imputation.

[0103] Example 2: Genotyping by sequencing Genotyping assays by sequencing were successfully performed on 223,266 samples, each assessed for coverage at the design site. Call rates are the percentage of sites with actionable genotypes. Figure 2 shows an average call rate of 98.9%, with 99.3% of samples having call rates of 95% or higher.

[0104] In addition to those described herein, various modifications of the described subject matter will be apparent to those skilled in the art from the foregoing description. Such modifications are intended to fall within the scope of the appended claims. Each reference cited in this application (including, but not limited to, journal articles, U.S. and non-U.S. patents, patent application publications, international patent application publications, gene bank accession numbers, etc.) is incorporated herein by reference in its entirety.

Claims

1. 1. A method for producing a nucleic acid probe for genotyping by sequencing, comprising: a) selecting a plurality of directly observed genetic variants for capture by said nucleic acid probe; b) filtering out low-confidence variants from the plurality of directly observed genetic variants, thereby creating a filtered plurality of directly observed genetic variants; c) phasing the filtered plurality of directly observed genetic variants; d) identifying, for each variant among the filtered plurality of directly observed genetic variants, the presence or absence of one or more proxy variants; e) selecting a plurality of candidate regions of genomic DNA comprising the filtered plurality of directly observed genetic variants, each candidate region of genomic DNA comprising about 25 to about 150 bases and comprising at least one variant among the filtered plurality of directly observed genetic variants; f) calculating for each candidate region of genomic DNA a quality score that estimates the capture efficiency and alignment success of the probe; g) calculating a probe score for each candidate region of genomic DNA by multiplying the quality score by the number of variants captured by the candidate region of genomic DNA, where the number of variants captured by the candidate region of genomic DNA is the sum of the number of directly observed variants captured by the candidate region of genomic DNA and the number of corresponding proxy variants in different candidate regions of genomic DNA; h) selecting one or more candidate regions of genomic DNA having the highest probe scores for inclusion in a final set of regions of genomic DNA; i) repeating steps g) and h) for candidate regions of unselected genomic DNA for inclusion in a final set of regions of genomic DNA, wherein the number of variants in the candidate regions of unselected genomic DNA is the sum of 1) the number of directly observed variants in the candidate regions of unselected genomic DNA, excluding all directly observed variants in the selected regions of genomic DNA, and 2) the number of corresponding proxy variants in different candidate regions of genomic DNA, excluding all proxy variants corresponding to the directly observed variants in the selected regions of genomic DNA, and steps g) and h) are repeated until a maximum number of regions of genomic DNA are selected; and j) generating a set of nucleic acid probes complementary to said nucleic acid sequences of each of said genomic regions in said final set of regions of genomic DNA. The method comprising:

2. 2. The method of claim 1, wherein the plurality of directly observed genetic variants is selected from a database of genome-wide association of genetic variants, a database of pharmacogenetic association of genetic variants, a database comprising genetic variants in all mitochondrial chromosomes, and / or a database of genetic variants in microarrays, or any combination thereof.

3. The squared correlation with the trait has a p-value of ≦10 -9 if a variant in the genome-wide association database of said genetic variants is retained in said plurality of directly observed genetic variants, and the squared association with the trait has a p-value > 10 -9 3. The method of claim 2, wherein a variant in a genome-wide association database of genetic variants is excluded from the plurality of directly observed genetic variants if

4. 3. The method of claim 2, wherein the database of genetic variants in the microarray comprises genetic variants in the HLA region of chromosome 6, the Y chromosome, two KIR regions on chromosome 19, and pseudoautosomal regions 1 and 2 (Par1 and Par2) on the X chromosome.

5. 5. The method of any one of claims 1 to 4, wherein multi-allelic variants are converted into one or more sets of bi-allelic variants.

6. 6. The method of any one of claims 1-5, wherein eliminating low confidence variants from the plurality of directly observed genetic variants comprises eliminating all variants having a minor allele frequency (MAF) below a desired threshold.

7. The method of claim 6 , wherein the desired threshold is 1%.

8. 8. The method of any one of claims 1-7, wherein eliminating low confidence variants from the plurality of directly observed genetic variants comprises eliminating all variants with missing data above a desired threshold.

9. The method of claim 8 , wherein the desired threshold is 2%.

10. 10. The method of any one of claims 1-9, wherein a variant among the filtered plurality of directly observed genetic variants has a corresponding proxy variant in another candidate region of genomic DNA when the directly observed genetic variant and the proxy variant are within 1 MB of each other and if the linkage disequilibrium between the two variants has a squared correlation of at least 0.2, at least 0.5, at least 0.8, at least 0.9, or at least 1.0 using an r2 measure of linkage disequilibrium.

11. The method of any one of claims 1 to 10, wherein the multiple candidate regions of the genomic DNA are divided into separate analysis groups, whereby each chromosome is a separate analysis group.

12. The method of any one of claims 1 to 11, wherein each candidate region of genomic DNA comprises from about 120 to about 125 bases.

13. 13. The method of any one of claims 1 to 12, wherein the plurality of candidate regions of genomic DNA comprises from about 5 million to about 50 million variants.

14. 14. The method of any one of claims 1 to 13, wherein the plurality of candidate regions of genomic DNA in total comprise from about 1 million to about 100 million base pairs, from about 5 million to about 75 million base pairs, from about 10 million to about 50 million base pairs, or from about 20 million to about 40 million base pairs.

15. 15. The method of any one of claims 1 to 14, wherein calculating the quality score comprises determining a component score for each of a mappability metric, an insertion-deletion mutation metric, and a classification metric for the candidate region of the genomic DNA, and wherein the quality score is a product of multiplications of each of the component scores.

16. The component scores of the mappability metric are exp(10×UmapMRM i -9), where UmapMRM i 16. The method of claim 15, wherein x is the multi-read mappability metric for variant position i within the candidate region of genomic DNA.

17. The insertion-deletion mutation metric is a measure of the presence or absence of an insertion or deletion of a base in the candidate region of the genomic DNA, and the component score of the insertion-deletion mutation is exp(SV score i ) where, if the variant position i is not linked to an insertion-deletion mutation or is linked to an insertion-deletion mutation of less than 5 bases, the SV score i is 2, and when the variant position i is linked to an insertion-deletion mutation of 5 to 10 bases, the SV score iis 1, and if the variant position i is linked to an insertion-deletion mutation of more than 10 bases, the SV score i The method of claim 16 , wherein:

18. The classification metric for the candidate region of the genomic DNA includes a first category, a second category, a third category, and a fourth category, and a first component score of the classification metric is exp(Region_score i ), whereby variant position i in the first category is scored as 0, variant position i in the second category is scored as 1, variant position i in the third category is scored as 1.6, and variant position i in the fourth category is scored as 2, and a second component score of the classification metric is (1+1.2(min(dist2category1 i , 60) / 60)), where dist2category1 i is the minimum absolute distance from the variant position i to a region in the first category, and the third component score of the classification metric is (1+1.2(min(dist2category2 i , 60) / 60)), where dist2category2 i The method of claim 15 , wherein x is the minimum absolute distance from the variant position i to a region of the second category.

19. selecting one or more candidate regions of said genomic DNA having the highest probe scores, identifying said candidate regions having three or more variants and having the highest probe scores; identifying said candidate regions having the highest probe scores and including only a subset of said regions with three or more variants; wherein all candidate regions are excluded from the plurality of candidate regions of the genomic DNA that include only a subset of the regions with three or more variants that have a probe score lower than the highest probe score of the candidate regions with three or more variants.

20. 1. A method for genotyping a DNA sample by sequencing, comprising: a) hybridizing a set of nucleic acid probes produced according to any one of claims 1 to 19 to said DNA sample to produce probe-hybridized genomic DNA; b) sequencing the genomic DNA hybridized to the probes to generate a plurality of sequencing reads; c) mapping the plurality of sequencing reads to a reference genome; d) calling the directly observed variants present in the mapped sequencing reads; and e) imputing unobserved variants from unsequenced regions of genomic DNA, thereby establishing the genotype of the sample DNA.

21. 1. A method for genotyping a DNA sample by sequencing using a set of nucleic acid probes, comprising: a) selecting a plurality of regions of genomic DNA from said DNA sample that contain a plurality of directly observed genetic variants; b) identifying the set of nucleic acid probes for hybridization to multiple regions of the selected genomic DNA; c) hybridizing the set of nucleic acid probes to the DNA sample to produce probe-hybridized genomic DNA; d) sequencing the genomic DNA hybridized to the probes to generate a plurality of sequencing reads; e) mapping the plurality of sequencing reads to a reference genome; f) calling the directly observed variants present in the mapped sequencing reads; and g) imputing unobserved variants from unsequenced regions of genomic DNA, thereby establishing the genotype of the sample DNA.

22. 1. A system comprising a data processor having a memory connected thereto, the memory comprising: selecting a plurality of regions of genomic DNA from the DNA sample that contain a plurality of directly observed genetic variants; identifying a set of nucleic acid probes for hybridization to a plurality of regions of said selected genomic DNA and sending instructions to an oligonucleotide synthesizer for synthesis of said set of nucleic acid probes; receiving a plurality of sequencing reads from a DNA sequencing device generated from sequencing a product of genomic DNA hybridized to the probes upon hybridization of the set of nucleic acid probes to a DNA sample; mapping the plurality of sequencing reads to a reference genome; calling the directly observed variants present in the mapped sequencing reads; and The system further comprises a program including instructions for imputing unobserved variants from unsequenced regions of genomic DNA, thereby establishing a genotype of the sample DNA.