Analysis of gene expression data using transitive directed graphs

University Of Memphis, Memphis TN

Investigators

Abstract

Recent advances in biotechnology have facilitated the simultaneous measurement of response of thousands of genes, and in effect enabled comparative studies that can identify which drugs are most effective and have fewest side effects at the genetic level. To obtain true response patterns of genes from these studies, researchers have often utilized a conservatively large number of samples for each treatment, increasing the cost significantly for studies with many drugs. Efforts aimed to reduce sample size (i.e. the number of samples in experiments) are challenged by the difficulty to articulate relationships between sample size and true patterns of gene response to treatments. Intellectual merit: This project will develop new methods that elucidate the relationships between true patterns of gene response to treatments and sample size. First, the use of directed graphs to represent gene response patterns make it easier for researchers to distinguish and visualize subtle gene-expression differences resulting from similar drugs. Further, properties of these graphs are exploited to isolate false patterns from true ones, enabling accurate predictions of true patterns of gene response even in cases where there are few samples. Accurate prediction of true response patterns of genes further leads to accurate prediction of gene function. Consequently, this work will pave the way for cost-effective experimental designs of comparative studies with many drugs. These results also help develop new methods that improve clustering analysis of gene response patterns. Unlike prominent conventional methods that treat all observed patterns as true, the advantage of this approach lies in the ability to quantify the degree to which each observed pattern could be true. This advantage will ultimately result in more accurate determination of groups of genes that cooperate in same pathways or share similar functions. Broader Impacts: This project will generate the materials that promote interdisciplinary teaching, training and research beyond the standard curriculum. Additionally, these materials will be refined and incorporated seamlessly into two separate events: the Computer Science Open House and the Programming Challenge, which have been organized annually by the PI for the past several years. Targeting high school students in the Memphis/Mid-South areas, these events aim to raise awareness, attract talent and foster interest in computing and computational sciences.