Systems and methods for haplotyping
The invention relates to methods for determining a haplotype for an organism by using a system for transforming SNP alleles found in sequence fragments into vertices in a graph with edges connecting vertices for alleles that appear together in a sequence fragment. A community detection operation can be used to infer the haplotype from the graph. The system may produce a report that includes the haplotype of the SNPs found in the genome of that organism.
1. A method for identifying haplotypes in a genome, the method comprising:
obtaining a plurality of sequence fragments from a genome of an organism;
transforming, using a processor coupled to a memory subsystem, the sequence fragments into a graph comprising a vertex for each allele of each of a plurality of SNPs found in the plurality of sequence fragments and an edge for each pair of the alleles that are found in one of the fragments;
for each pair of the plurality of SNPs for which alleles are found in one of the fragments, determine a most-supported phase for alleles of that pair of SNPs and remove any edge from the graph representing a less-supported phase for the alleles of that pair of SNPs; and
apply a community detection operation to the largest contiguous component of the graph remaining after the edge removal to assign each vertex of that component to a haplotype,
wherein the haplotype covers at least 85% of a chromosome.
2. The method of claim 1 , wherein transforming the sequence fragments into a graph comprises:
creating the vertices to use index-free adjacency, wherein each vertex includes one pointer for each connected vertex to which that vertex is connected by an edge,
wherein each pointer identities a location of the connected vertex, and
wherein identifying the location of a connected vertex by a pointer includes identifying a physical location in the memory subsystem where the connected vertex is stored.
3. The method of claim 1 , wherein the applying the community detection comprises finding a maximum-likelihood assignment of vertices to one or more blocks, wherein the probability of the graph given the assignment is maximized.
4. The method of claim 3 , wherein finding the maximum likelihood assignment includes:
(i) assigning each node to one of the two blocks;
(ii) calculating the probability of the graph given the assignment;
(iii) making a change to the assignment;
(iv) repeating steps (ii) and (iii) to create a chain of assignments; and
(v) selecting from the chain the assignment wherein the probability of the graph given that assignment is optimized.
5. The method of claim 4 , wherein repeating steps (ii) and (iii) includes modifying the block membership of each vertex in a random fashion and accepting or rejecting the modification according to a change in the probability of the assignment associated with that modification.
6. The method of claim 3 , wherein obtaining the plurality of sequence fragments from a genome of an organism comprises sequencing nucleic acid from a sample from the organism.
7. The method of claim 3 , wherein the organism is a patient and the plurality of sequence fragments are obtained by sequencing nucleic acid from the patient.
8. The method of claim 7 , further comprising producing a report showing the haplotype for the patient.
9. The method of claim 1 , wherein each of the plurality of sequence fragments comprises either a sequence read or a pair of paired-end sequence reads.
10. The method of claim 1 , wherein at least some of the SNPs correspond to multiple alleles.
11. A method for identifying haplotypes in a genome, the method comprising:
obtaining a plurality of sequence fragments generated by sequencing nucleic acid from a genome of a patient;
creating, using a computer system comprising a processor coupled to a memory subsystem, a graph comprising a vertex for each allele of each of a plurality of SNPs found in the plurality of sequence fragments and an edge for each pair of the alleles that are found in one of the fragments, wherein the graph uses pointers to identify a physical location in the memory subsystem where each vertex is stored;
determining, for each pair of the plurality of SNPs for which alleles are found in one of the fragments, a best-supported phase for alleles of that pair of SNPs and remove at least one edge from the graph representing a less-supported phase for alleles of that pair of SNPs;
finding a maximum likelihood assignment of vertices to one or more blocks wherein the probability of the graph given the assignment is maximized, thereby assigning each allele to a haplotype; and
producing a report showing the haplotype for the patient,
wherein the haplotype covers at least 85% of a chromosome.
12. The method of claim 11 , wherein finding the maximum likelihood assignment includes:
(i) assigning each vertex to one of the two blocks;
(ii) calculating the probability of the graph given the assignment;
(iii) making a change to the assignment;
(iv) repeating steps (ii) and (iii) to create a chain of assignments; and
(v) selecting the maximum likelihood assignment from the chain of assignments.
13. The method of claim 12 , wherein repeating steps (ii) and (iii) defines a Monte Carlo Markov chain by modifying the block membership of each vertex in a random fashion and accepting or rejecting the modification according to a change in the probability of the assignment associated with that modification.
14. The method of claim 13 , wherein obtaining the plurality of sequence fragments includes sequencing nucleic acid from a sample from the patient.
15. The method of claim 13 , wherein each of the plurality of sequence fragments comprises either a sequence read or a pair of paired-end sequence reads.
16. A system for identifying haplotypes in a genome, the system comprising a processor coupled to a memory subsystem, wherein the system is operable to:
obtain a plurality of sequence reads generated by sequencing nucleic acid from a genome of a patient;
create a graph comprising a vertex for each allele of each of a plurality of SNPs found in the plurality of sequence fragments and an edge for each subset of the alleles that are found in one of the fragments;
determine, for each pair of the plurality of SNPs for which alleles are found in one of the fragments, a best-supported phase for alleles of that pair of SNPs and remove at least one edge from the graph representing a less-supported phase for alleles of that pair of SNPs;
find an optimal assignment of vertices to one or more blocks by a community detection operation, thereby assigning each allele to a haplotype; and
produce a report showing the haplotype for the patient,
wherein the haplotype covers at least 85% of a chromosome.
17. The system of claim 16 , wherein finding the optimal assignment includes:
(i) assigning each vertex to one of the two blocks;
(ii) calculating the probability of the graph given the assignment;
(iii) making a change to the assignment;
(iv) repeating steps (ii) and (iii) to create a chain of assignments; and
(v) selecting maximum likelihood assignment from the chain.
18. The system of claim 17 , wherein repeating steps (ii) and (iii) defines a Monte Carlo Markov chain by modifying the block membership of each vertex in a random fashion and accepting or rejecting the modification according to a change in the probability of the assignment associated with that modification.
19. The method of claim 10 , wherein the graph includes at least two paths each corresponding to a haplotype that itself includes multiple alleles.
20. The method of claim 11 , wherein at least some of the SNPs correspond to multiple alleles.
21. The method of claim 20 , wherein the graph includes at least two paths each corresponding to a haplotype that itself includes multiple alleles.
22. The system of claim 16 , wherein at least some of the SNPs correspond to multiple alleles.
23. The system of claim 22 , wherein the graph includes at least two paths each corresponding to a haplotype that itself includes multiple alleles.