The New Jersey Institute of Technology's
Electronic Theses & Dissertations Project

Title:	Cancer risk prediction with next generation sequencing data using machine learning
Author:	Patel, Nihir
View Online:	njit-etd2015-010 (xi, 33 pages ~ 9.4 MB pdf)
Department:	Department of Computer Science
Degree:	Master of Science
Program:	Bioinformatics
Document Type:	Thesis
Advisory Committee:	Roshan, Usman W. (Committee chair) Wang, Jason T. L. (Committee member) Wei, Zhi (Committee member)
Date:	2015-01
Keywords:	Cancer risk prediction Next generation sequencing Machine learning
Availability:	Unrestricted
Abstract:	The use of computational biology for next generation sequencing (NGS) analysis is rapidly increasing in genomics research. However, the effectiveness of NGS data to predict disease abundance is yet unclear. This research investigates the problem in the whole exome NGS data of the chronic lymphocytic leukemia (CLL) available at dbGaP. Initially, raw reads from samples are aligned to the human reference genome using burrows wheeler aligner. From the samples, structural variants, namely, Single Nucleotide Polymorphism (SNP) and Insertion Deletion (INDEL) are identified and are filtered using SAMtools as well as with Genome Analyzer Tool Kit (GATK). Subsequently, the variants are encoded and feature selection is performed with the Pearson correlation coefficient (PCC) and the chi-square 2-df statistical test. Finally, 90:10 cross validation is performed by applying the support vector machine algorithm on sets of top selected features. It is found that the variants detected with SAMtools and GATK achieve similar prediction accuracies. It is also noted that the features that are ranked with the PCC yield better accuracy than the chi-square test. In all of the analyses, the SNPs are identified to have superior accuracy as compared to the INDELs or the full dataset. Later, an exome capture kit is introduced for analysis. The SNPs, ranked with the PCC, along with the exome capture kit yield prediction accuracy of 85.1 % and area under curve of 0.94. Overall, this study shows the effective application of the machine learning methods and the strength of the NGS data for the CLL risk prediction.