Predicting the Lethal Phenotype of the Knockout Mouse by Integrating Comprehensive Genomic Data

• Activity Number: 253
• Type: Contributed
• Date/Time: Monday, July 30, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Robyn Ball*+ and Yuan Yuan and Yanxun Xu and Jianfeng Xu and Han Liang
• Companies: and Baylor College of Medicine and Rice University/MD Anderson Cancer Center and Alliance Data and MD Anderson Cancer Center
• Keywords: random forest ; machine learning ; knockout mouse

Abstract

The mouse is widely used as a model organism. A conventional approach in mouse genetics research is to investigate the phenotypic consequence of knocking out a gene. One commonly-studied phenotype is lethality, a binary outcome. Using data collected from the Mouse Genome Informatics (MGI) Database, we compared the accuracy of three genomic classifiers for predicting lethality. Each of logistic regression, support vector machine, and random forest was used for classification, with feature-selection carried out by LASSO in each case. Error rates were estimated with cross validation. We found that the random forest classifier almost always outperformed the other methods, with an estimated accuracy of ~74%.