EMINIM is a software for imputation of unobserved genotypes using a set of reference haplotype panel at a higher-density SNP set such as HapMap, and lower-density genotypes of a target individual using such as genotyping arrays. Our method is adaptive in the sense that it learns transtition/emission parameters of hidden Markov model using EM algorithm with standard Poisson point process for modeling recombination. Our method is memory-efficient by efficiently utilizing silent states in the HMM. EMINIM provides interfaces both for human and inbred model organisms such as laboratory mice. Our method is demonstrated to have ten times smaller errors than the previous imputation method designed for inbred mouse for high-confidence genotypes.
:: MORE INFORMATION
J Comput Biol. 2010 Mar;17(3):547-60.
EMINIM: an adaptive and memory-efficient algorithm for genotype imputation.
Kang HM, Zaitlen NA, Eskin E.