Abstract
Haplotypes are a kind of powerful information that is helpful in gene candidate studies because of inheritance characteristics. However, in order to get the haplotype information, in vitro methods cost lots of time and money, it is helpful to infer haplotypes using in Silico methods. Because the haplotype inference is a NP-Hard problem, both the accuracy and computational time are important issues. In this thesis, we take into account of the normalized mutual information in the parsimonious tree-grow methods that show very good performance on haplotype inference problems. And we improve the inference accuracy rate most to 2.62 percent on APOE gene dataset which just spend about 0.001 more seconds than original parsimonious tree-grow method. We also have the highest to 95.2% accuracy rate on β2AR gene data in comparison to previous approaches.