Monte Carlo multipoint linkage analysis

Journal Article · Thu Sep 01 04:00:00 EDT 1994 · American Journal of Human Genetics

OSTI ID:133422

Lin, S ^[1]; Wijsman, E M ^[2]

---

1. Univ. of California, Berkeley, CA (United States)
2. Univ. of Washington, Seattle, WA (United States)

Multipoint lod scores can be formidable or even impossible to obtain in situations where there are large numbers of markers in the analysis, and in particular, when there are large numbers of alleles per marker. In order to circumvent these computational problems, Markov chain Monte Carlo (MCMC) methods have recently been explored as a method of estimating the likelihood function and corresponding likelihood ratios. These methods are based on the generation of a series of dependent samples drawn from the desired equilibrium distribution, which is achieved if all legal genotypic configurations communicate. We present methods which implement MCMC methods for use in multipoint linkage analysis. This implementation builds on several recent advances in these methods as applied to pedigree analysis. First, because irreducibility is not guaranteed for multi-allelic markers, in addition to the true penetrance matrix for the real chain, we introduce at least one modified penetrance matrix for constructing each irreducible companion chain defined on an augmented space. The states of the real chain and the companion chain(s) are switched with some probability after each scan thereby guaranteeing irreducibility for the whole process. Second, we increase efficiency of estimation by defining a smaller augmented space in the companion chain(s) after finding all non-communicating sets of the genotypic configuration space. Finally, we further increase efficiency by introducing parameters to improve the switching rate between chains. We use data on a dominant disorder and multi-allelic markers to show that the estimates of the lod scores obtained with MCMC methods are virtually identical to those obtained with exact methods with LINKAGE. These results demonstrate that MCMC methods can be used to perform multipoint linkage analyses with multi-allelic markers, and thus provides a means of performing such analyses in situations where exact computation is impractical.

• OSTI ID: 133422
• Report Number(s): CONF-941009--
• Journal Information: American Journal of Human Genetics, Journal Name: American Journal of Human Genetics Journal Issue: Suppl.3 Vol. 55; ISSN AJHGAG; ISSN 0002-9297
• Country of Publication: United States
• Language: English

Related Subjects

55 BIOLOGY AND MEDICINE
BASIC STUDIES
99 GENERAL AND MISCELLANEOUS
BIOLOGICAL MARKERS
COMPUTER CODES
EFFICIENCY
GENES
GENETICS
GENOTYPE
HEREDITARY DISEASES
HUMAN POPULATIONS
MONTE CARLO METHOD
RISK ASSESSMENT
STATISTICS