UC Berkeley

In this work, I develop computational tools focused around the utilization of DNA sequence data to address questions relative to forensic science, medical genetics, human evolution and ancient DNA.

First, I compute the theoretical probability that two individual profiles match by chance at two loci in a subdivided population. This question is of particular interest in forensic science, where DNA evidence has become a widespread tool of investigation and criminal conviction. I find that the effect of ignoring population subdivision can be unfavorable to

the defendant, but that the two loci can essentially be treated as unlinked.

Second, I develop a method to identify genes that are interacting, or in epistasis, to produce a particular phenotype. Determining interacting genes is indeed of particular relevance in medical genetics to help map disease genes. I validate the method with simulations and

demonstrate an improved performance over existing approaches. I also apply the method to recently available genomic data from domesticated dogs, identifying genes in epistastis for the hair length phenotype - thus representing candidate genes for functional validation.

Third, I use a summary statistic of DNA sequences, the site frequency spectrum, to estimate parameters of recent human history, and to characterize the potential event of admixture between Neanderthals and humans. I find evidence for recent gene flow between

Neanderthals and Europeans, and to a lesser extent between Neanderthals and Africans.

Finally, I develop a likelihood method to jointly estimate the age and selection coefficient of an identified mutation, along with the population size, by using time serial samples. Such datasets are widespread in the fields of ancient DNA as well as experimental and viral

evolution. I validate the method through simulations. I re-analyze a recent dataset for a locus coding for the distribution of black pigmentation in horses - and estimate that the allele far predates domestication, arising between 20,000 and 13,000 years ago.