Systems and methods for inferring genetic ancestry from low-coverage genomic data

1. A computer-implemented method for a genetic screening procedure for a patient of a healthcare provider, at least a portion of the method being performed by a computing device comprising at least one processor, the method comprising:

deriving genetic data for the patient using a low-coverage next-generation sequencing procedure, the genomic data comprising raw sequencing data and a plurality of loci corresponding to genetic markers in a genome of the patient;

identifying, based on the genomic data of the patient, a genetic reference dataset that (i) comprises reference genomic reads for subjects other than the patient, and that (ii) comprises, for each subject, dosages at reference loci corresponding to the plurality of loci in the genomic data of the patient;

generating a reference matrix representing the genetic reference dataset, wherein the reference matrix comprises a first dimension corresponding to the subjects and a second dimension corresponding to the reference loci, and wherein each cell of the reference matrix indicates a dosage for a subject at a locus in the reference loci;

generating, based on the reference matrix, an ancestral attribution matrix attributing descent from one or more of a plurality of proposed ancestral populations to each of the subjects;

deriving, based on the ancestral attribution matrix, an ancestral attribution vector of the patient that comprises a proportion of descent for each subject attributable to each of the plurality of proposed ancestral populations, wherein deriving the ancestral attribution vector of the patient comprises generating a plurality of simulated ancestral attribution vectors, and identifying one or more of the simulated ancestral attribution vectors to attribute an admixture of populations to the patient by a decision making technique with a K-dimensional tree; and

determining a genetic ancestry of the patient based on the identified one or more ancestral attribution vectors.

2. The computer-implemented method of claim 1 , further comprising applying a sequencing procedure to a biological sample from the patient to derive the genomic data, wherein the sequencing procedure is a high-throughput sequencing procedure.

3. The computer-implemented method of claim 2 , wherein:

the high-throughput sequencing procedure comprises a targeted sequencing procedure; and

the genomic data comprise off-target genomic reads.

4. The computer-implemented method of claim 2 , wherein the genomic data comprise information indicating a likelihood of correctness for genomic reads in the genomic data of the patient.

5. The computer-implemented method of claim 1 , wherein identifying the genetic reference dataset comprises:

identifying missing dosage data within the genetic reference dataset, wherein the missing dosage data comprises dosage information present at corresponding reference loci for some but not all of the subjects; and

imputing the missing dosage data at a given locus based at least in part on observed correlations between the reference genetic reads for subjects without dosage information at the given locus and reference genetic reads for subjects with dosage information at the given locus.

6. The computer-implemented method of claim 1 , wherein identifying the genetic reference dataset comprises imputing dosage data to the genetic reference dataset based at least in part on an additional genetic dataset.

7. The computer-implemented method of claim 1 , wherein a subset of dosages within the reference matrix comprises probabilistic and continuous dosage values.

8. The computer-implemented method of claim 1 , wherein generating the ancestral attribution matrix comprises modeling each reference genome as being derived in specified proportions from each of the plurality of proposed ancestral populations.

9. The computer-implemented method of claim 8 , wherein generating the ancestral attribution matrix based on the reference matrix comprises constraining the generation of the ancestral attribution matrix such that:

each value representing a proportion of descent from one of the plurality of proposed ancestral populations falls between zero and one; and

for reference genetic reads of each subject, a sum of values representing proportions of descent from each of the plurality of proposed ancestral populations equals one.

10. The computer-implemented method of claim 1 , wherein generating the ancestral attribution matrix comprises penalizing non-sparse ancestral attribution matrix candidates to a predetermined degree.

11. The computer-implemented method of claim 1 , further comprising deriving an ancestral alternate reads matrix based on the ancestral attribution matrix, wherein deriving the ancestral alternate reads matrix comprises using a least squares approximation based on the ancestral attribution matrix and the reference matrix.

12. The computer-implemented method of claim 1 , wherein deriving the ancestral attribution vector comprises approximating the ancestral attribution vector using a least squares approximation based on an ancestral alternate reads matrix and the ancestral attribution matrix, the ancestral alternative reads matrix being based on the ancestral attribution matrix and the reference matrix.

13. The computer-implemented method of claim 1 , wherein deriving the ancestral attribution vector comprises approximating the ancestral attribution vector such that:

the ancestral attribution vector is constrained to including values between zero and one; and

a summation of values within the ancestral attribution vector is constrained to be one.

14. The computer-implemented method of claim 1 , further comprising generating a plurality of reference clusters from the reference genetic reads located within a multi-dimensional space, wherein each dimension within the multi-dimensional space corresponds to the proportion of descent from one of the plurality of proposed ancestral populations.

15. The computer-implemented method of claim 14 , further comprising iteratively merging cluster pairs within the plurality of reference clusters until a distance between a closest cluster pair within the plurality of reference cluster exceeds a predetermined threshold.

16. The computer-implemented method of claim 14 , further comprising defining a plurality of dines connecting each possible pair of reference clusters within the plurality of reference clusters, each cline within the plurality of dines describing a line connecting a pair of reference clusters within the plurality of reference clusters and representing a population composed of an admixture of two reference populations.

17. The computer-implemented method of claim 16 , wherein determining the genetic ancestry of the patient subject further comprises determining a membership of the patient to a dine within the plurality of dines based at least in part on a probability that the ancestral attribution vector pertains to the dine.

18. The computer-implemented method of claim 1 , wherein the sequencing procedure has a sequencing depth of less than 0.1x.

19. The computer-implemented method of claim 1 , wherein the genetic screening procedure is a carrier screening procedure or an inherited cancer screening procedure.

20. A computing system for inferring genetic ancestry from low-coverage genomic data for a genetic screening procedure, the computing system comprising:

at least one processor;

a receiving module, stored in computer-readable non-transitory memory, that receives, via the at least one processor, through an input device or a communication interface, low-coverage genomic data of a patient, the genomic data comprising raw sequencing data and a plurality of loci corresponding to genetic markers;

an identification module, stored in the computer-readable non-transitory memory, that identifies, via the at least one processor, based on the genomic data of the patient, a genetic reference dataset that (i) comprises reference genomic reads for subjects other than the patient, and that (ii) comprises, for each subject, dosages at reference loci corresponding to the plurality of loci in the genomic data of the patient;

a generation module, stored in the computer-readable non-transitory memory, that generates, via the at least one processor, a reference matrix representing the genetic reference dataset, wherein the reference matrix comprises a first dimension corresponding to the subjects and a second dimension corresponding to the reference loci, and wherein each cell of the reference matrix indicates a dosage for a subject at a locus in the reference loci;

a decomposition module, stored in the computer-readable non-transitory memory, that generates, via the at least one processor, based on the reference matrix, an ancestral attribution matrix attributing descent from one or more of a plurality of proposed ancestral populations to each of the subjects;

a derivation module, stored in the computer-readable non-transitory memory, that derives, via the at least one processor, based on the ancestral attribution matrix, an ancestral attribution vector of the patient that comprises a proportion of descent for each subject attributable to each of the plurality of proposed ancestral populations, wherein deriving the ancestral attribution vector of the patient comprises generating a plurality of simulated ancestral attribution vectors, and identifying one or more of the simulated ancestral attribution vectors to attribute an admixture of populations to the patient by a decision making technique with a K-dimensional tree; and

a determination module, stored in the computer-readable non-transitory memory, that determines, via the at least one processor, the genetic ancestry of the patient based on the identified one or more ancestral attribution vectors.

21. A non-transitory computer-readable medium comprising one or more computer-readable instructions that, when executed by at least one processor of a computing device, cause the computing device to:

receive, through an input device or a communication interface of the computing device, low-coverage genomic data of a patient identified as having an ethnicity in the computing device, the genomic data comprising raw sequencing data and a plurality of loci corresponding to genetic markers;

identify, based on the genomic data, a genetic reference dataset that (i) comprises reference genomic reads for subjects other than the patient, and that (ii) comprises, for each subject, dosages at reference loci corresponding to the plurality of loci in the genomic data of the patient;

generate a reference matrix representing the genetic reference dataset, wherein the reference matrix includes a first dimension corresponding to the subjects and a second dimension corresponding to the reference loci, and wherein each cell of the reference matrix indicates a dosage for a subject at a locus in the reference loci;

generate, based on the reference matrix, an ancestral attribution matrix attributing descent from one or more of a plurality of proposed ancestral populations to each of the subjects;

derive, based on the ancestral attribution matrix, an ancestral attribution vector of the patient that comprises a proportion of descent for each subject attributable to each of the plurality of proposed ancestral populations, wherein deriving the ancestral attribution vector of the patient comprises generating a plurality of simulated ancestral attribution vectors, and identifying one or more of the simulated ancestral attribution vectors to attribute an admixture of populations to the patient by a decision making technique with a K-dimensional tree; and

determine the genetic ancestry of the patient based on the identified one or more ancestral attribution vectors.