Trio-based phasing using a dynamic Bayesian network
Performing trio-based phasing includes: obtaining a set of preliminary phased haplotype data of an individual; establishing a dynamic Bayesian network based at least in part on the set of preliminary phased haplotype data of the individual and phased haplotype data of at least one parent of the individual; and determining, based on the dynamic Bayesian network, a set of refined haplotype data of the individual.
1. A system for performing trio-based phasing, comprising:
one or more processors configured to:
obtain a set of preliminary phased haplotype data of an individual;
establish a dynamic Bayesian network based at least in part on the set of preliminary phased haplotype data of the individual and phased haplotype data of at least one parent of the individual; and
determine, based on the dynamic Bayesian network, a set of refined haplotype data of the individual; and
one or more memories coupled with the one or more processors, configured to provide the one or more processors with instructions.
2. The system of claim 1 , wherein the dynamic Bayesian network is established to model observed alleles of two haplotypes, hidden true alleles of the individual's parental haplotype, and transmission of the parental haplotype to the individual.
3. The system of claim 1 , wherein to determine the set of refined haplotype data of the individual includes to:
pre-compute emission probabilities associated with a position on a chromosome of the individual;
determine transition probabilities based at least in part on transition probabilities from a previous position;
search for likely settings of transition variables across a chromosome sequence; and
search for likely settings of emission variables given observed variables and transition variables.
4. The system of claim 1 , wherein genotype data at a particular site is represented using:
G 0 *,i ,G 1 *,i ε{0, 1} representing observed alleles for a first haplotype of the set of preliminary phased haplotype data and a second haplotype of the set of preliminary phased haplotype data;
H m *,i ,H p *,i ε{0, 1} representing hidden true alleles of the individual's maternal (m) and paternal (p) haplotypes; and
P *,i ε{m, p} representing a hidden binary phase indicator variable that is set to m whenever G 0 *,i corresponds to H m *,i and set to p whenever G 0 *,i corresponds to H p *,i .
5. The system of claim 4 , wherein determine, based on the dynamic Bayesian network, the set of refined haplotype data of the individual includes to maximize P(H m kid ,H p kid ,H m mom ,H p mom ,H m dad ,H p dad ) given H m i +H p i =G 0 i +G 1 i ∀iε{kid,mom,dad}.
6. The system of claim 1 , wherein a relationship between parental haplotype and haplotype of the individual is represented by T mom,i ,T dad,i ε{a, b}, where a indicates transmission of a parent's maternal haplotype to the child and b indicates transmission of a parent's paternal haplotype to the child.
7. The system of claim 1 , wherein the set of preliminary phased haplotype data is obtained using population-based phasing of unphased genotype data of the individual.
8. The system of claim 1 , wherein to determine the set of refined haplotype data of the individual includes using the dynamic Bayesian model to:
pre-compute emission probabilities associated with a position on a chromosome of the is individual;
determine transition probabilities based at least in part on transition probabilities from a previous position;
search for likely settings of transition variables across a chromosome sequence; and
search for likely settings of hidden variables given observed variables and transition variables.
9. A method of performing trio-based phasing, comprising:
obtaining a set of preliminary phased haplotype data of an individual;
establishing a dynamic Bayesian network based at least in part on the set of preliminary phased haplotype data of the individual and phased haplotype data of at least one parent of the individual; and
determining, based on the dynamic Bayesian network, a set of refined haplotype data of the individual.
10. The method of claim 9 , wherein the dynamic Bayesian network is established to model observed alleles of two haplotypes, hidden true alleles of the individual's parental haplotype, and transmission of the parental haplotype to the individual.
11. The method of claim 9 , wherein determining the set of refined haplotype data of the individual includes:
pre-computing emission probabilities associated with a position on a chromosome of the individual;
determining transition probabilities based at least in part on transition probabilities from a previous position;
searching for likely settings of transition variables across a chromosome sequence; and
searching for likely settings of emission variables given observed variables and transition variables.
12. The method of claim 9 , wherein genotype data at a particular site is represented using:
G 0 *,i ,G 1 *,i ε{0, 1} representing observed alleles for a first haplotype of the set of is preliminary phased haplotype data and a second haplotype of the set of preliminary phased haplotype data;
H m *,i ,H p *,i ε{0, 1} representing hidden true alleles of the individual's maternal (m) and paternal (p) haplotypes; and
P *,i ε{m, p} representing a hidden binary phase indicator variable that is set to m whenever G 0 *,i corresponds to H m *,i nd set to p whenever G 0 *,i corresponds to H p *,i .
13. The method of claim 12 , wherein determining, based on the dynamic Bayesian network, the set of refined haplotype data of the individual includes maximizing P(H m kid ,H p kid ,H m mom ,H p mom ,H m dad ,H p dad ) given H m i +H p i =G 0 i +G 1 i ∀iε{kid,mom,dad}.
14. The method of claim 9 , wherein a relationship between parental haplotype and haplotype of the individual is represented by T mom,i ,T dad,i ε{a, b}, where a indicates transmission of a parent's maternal haplotype to the child and b indicates transmission of a parent's paternal haplotype to the child.
15. The method of claim 9 , wherein the set of preliminary phased haplotype data is obtained using population-based phasing of unphased genotype data of the individual.
16. The method of claim 9 , wherein determining the set of refined haplotype data of the individual includes using the dynamic Bayesian model to:
pre-compute emission probabilities associated with a position on a chromosome of the individual;
determine transition probabilities based at least in part on transition probabilities from a previous position;
search for likely settings of transition variables across a chromosome sequence; and
search for likely settings of hidden variables given observed variables and transition variables.
17. A computer program product for performing trio-based phasing, the computer program product being embodied in a tangible non-transitory computer readable storage medium and comprising computer instructions for:
obtaining a set of preliminary phased haplotype data of an individual;
establishing a dynamic Bayesian network based at least in part on the set of preliminary is phased haplotype data of the individual and phased haplotype data of at least one parent of the individual; and
determining, based on the dynamic Bayesian network, a set of refined haplotype data of the individual.