No. 928

Title:

Identifying haplotype block structure by using ancestor-derived model and MDL principle

Author(s):

Fujisawa, Hironori (The Institute of Statistical Mathematics);
Isomura, Minoru (Genome Center, Japanese Foundationfor Cancer Research);
Eguchi, Shinto (The Institute of Statistical Mathematics);
Ushijima, Masaru (Genome Center, Japanese Foundationfor Cancer Research);
Miyata, Satoshi (Genome Center, Japanese Foundationfor Cancer Research);
Miki, Yoshio (Genome Center, Japanese Foundationfor Cancer Research);
Matsuura, Masaaki (Genome Center, Japanese Foundationfor Cancer Research)

Key words:

Ancestral haplotype; Dynamic programming algorithm; Hotspot sensitivity; Linkage disequilibrium; Low haplotype diversity; Probabilistic structure of recombination; Recombinant haplotype; Recombination hotspot; Single nucleotide polymorphism

Abstract:

    A method for identifying a haplotype block structure is proposed using an ancestorderived model and the MDL principle. The haplotype block structure is caused by the existence of ancestral haplotypes and recombination events. Using this idea, the ancestor-derived model is constructed. Furthermore, the whole statistical model is proposed as a modification of the ancestor-derived model to treat some nonstandard data. The haplotype block structure is identified using the optimal model selected by the MDL principle. The simulation study indicates that the proposed method is powerful from the viewpoint of hotspot sensitivity and robust to mutation except near the edge of a sequence. The proposed method was applied to two real data sets: the JFCR data and the 5q31 data of Daly et al. The analysis of the former data gave a clear explanation for a conventional association study of an antitumor drug. The analysis of the latter data presented a similar haplotype block structure to that in Daly et al.