Computer Science Department
McGregory Hall, 3rd Floor
13 Oak Drive
Hamilton, NY 13346
Charlotte Jablonski, Administrative Assistant
Start: Tuesday, April 21, 2015, 11:20 a.m.
End: Tuesday, April 21, 2015, 12:10 p.m.
Location: 329 McGregory
Join us for our next department tea! We will have two speakers, Farah Fouladi '15 and Sam Daulton '15. Each will speak about their final project from Professor Ay's course on Modeling of Biological Systems.
Lunch will be served following the talks.
Title: Predicting Cancer using Mutual Information-based Gene Association Networks
Speaker: Sam Daulton '15
Abstract: Machine-learning algorithms can be used on gene expression data as an alternative to traditional clinical methods for diagnosing cancer and determining a patient’s prognosis. A standardized, accurate approach to classifying and diagnosing colon cancer does not exist. Little is known about molecular alterations associated with the heterogeneity of the disease, and no molecular marker has been validated for clinical practice as a diagnostic or prognostic parameter. We propose a novel supervised classification algorithm, the NBC-A method, for predicting cancer from gene expression data with greater accuracy and for identifying genetic biomarkers. NBC-A improves the NBC method developed by Ay et al. in 2014 by constructing gene association networks using mutual information as the statistical metric instead of Pearson correlation. We expect NBC-A to produce transcriptional networks that are more indicative of the underlying biological pathways of colorectal cancer, leading to discovery of new biomarkers. NBC-A yields higher classification accuracies than many traditional classifiers including Support Vector Machine (SVM), k-Nearest Neighbors (kNN), and Naïve Bayes (NB) on the tested colon cancer dataset. In addition, NBC-A outperforms all other methods tested (NBC, SVM, kNN, NB, and Random Forest (RF)) on a lung cancer dataset.
Speaker: Farah Fouladi '15