Authors

Shengnan Cai

Type

Text

Type

Dissertation

Advisor

Hongshik Ahn | Ahn, Hongshik | Nancy Mendell | Stephen Finch | Sangjin Hong.

Date

2010-12-01

Keywords

Classification models, Cross validation, Next generation sequencing, SNP detection, Variable selection | Statistics

Department

Department of Applied Mathematics and Statistics

Language

en_US

Source

This work is sponsored by the Stony Brook University Graduate School in compliance with the requirements for completion of degree.

Identifier

http://hdl.handle.net/11401/70951

Publisher

The Graduate School, Stony Brook University: Stony Brook, NY.

Format

application/pdf

Abstract

Variations in DNA sequences of humans have a strong association with many diseases. Single Nucleotide Polymorphism (SNP) is the most common type of DNA variations. Our research is to detect SNPs from the data generated by Polymerase Chain Reaction (PCR) and next generation sequencing methods. In the first part of the study, we had a relatively small data set with fewer known SNPs as the training data. We developed a classification model based on the cross validation method. From the first part of the research, we gained knowledge of the properties of the data. In the next phase, we obtained a much larger data set with a much larger group of known SNPs. We developed eight measures for every genetic position with these data. Using these eight measures as the predictor variables, we applied several classification methods such as Random Forest (RF), Support Vector Machines (SVM), Single Decision Tree (ST) and Logistic Regression (LR); then used cross validation to evaluate these classification methods. By comparing the predictive accuracy, sensitivity and specificity, we found the best performing model for the data. To compare the performances of these models while the number of observations for each genetic position (cover depth) is small, we randomly drew out subsets from the whole data and applied these classification models. Variable selection is also used to our study. The result shows, SVM using the selected variables has a significant higher average accuracy than the other methods in general, but RF using the selected variables performs the best when the cover depth is as small as 20.

Share

COinS
 
 

To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.