Date of Graduation

2017

Document Type

Thesis

Degree Type

MS

College

Statler College of Engineering and Mineral Resources

Department

Lane Department of Computer Science and Electrical Engineering

Committee Chair

Donald Adjeroh

Committee Co-Chair

Jeremy Dawson

Committee Member

Tim Driscoll

Committee Member

YanFang Ye

Abstract

We study the problem of predicting human biogeographical ancestry using genomic data. While continental level ancestry prediction is relatively simple using genomic information, distinguishing between individuals from closely associated sub-populations (e.g., from the same continent) is still a difficult challenge. In particular, we focus on the case where the analysis is constrained to using single nucleotide polymorphisms (SNPs) from just one chromosome. We thus propose methods to construct ancestry informative SNP panels analyzing variants from a single chromosome, and evaluate the performance of such panels for both continental-level and sub-continental level ancestry prediction.;Efficient selection of ancestry informative SNPs is the key to successful ancestry prediction. The removal of redundant and noisy SNP features is essential prior to applying a learning algorithm. Here we propose two distinct methods of SNP selection: one is correlation-based SNP selection which uses a correlation metric to evaluate the usefulness of SNP features, while the other is random subspace projection based SNP selection which uses the learning algorithm itself to evaluate the worth of the SNP features. Correlation-based SNP selection approach can construct a small panel of useful SNPs for both continental level classification as well as binary classification of sub-populations. Unlike the correlation-based selection, random subspace projection based selection can construct efficient panel of SNP markers to address the difficult task of multinomial classification with multiple closely related sub-populations. We include results that demonstrate the performance of both methods, including comparison with other recently published related methods.

Download

Share

COinS