Seminars/Colloquia/Invited Talks

Seminars

David Kulp

Gene network inference from in-situ hybridization images, genotypes, and transcription profiling

PLACE:	Clark 314
EVENT:	CIS Seminar Series
DATE:	December 5, 2006
TIME:	1:00 - 2:00 PM

Abstract

I will speak on two topics concerned with the identification of related genes in common pathways. The first is about the identification of functionally coherent genes by spatial similarity from in-situ hybridizations in mouse brain. The Allen Brain Institute has performed gene-specific staining of slices of mouse brain -- one brain per gene -- which reveal the location of gene expression and its concentration. We hypothesized that similarities in spatial gene expression would reveal clusters of genes in common pathways. To address this question, we developed an end-to-end system for processing raw images and performing analysis. We use a non-linear image registration technique specifically adapted for mapping expression images to anatomical annotations and a method for extracting expression information. We then employ a bi-clustering technique over location and gene subsets and we relate the clustered patterns to Gene Ontology (GO) annotations.
In the second half of my talk, I will describe our work inferring causally correct pairs of regulator-target genes. In this data set, whole genome gene expression is measured using microarrays on more than 100 mice and each mouse was also finely genotyped. The key idea in our work is that the epistatic-like effect between genotype and gene expression of regulatory genes allows for the prediction of regulator-target pairs. Pairwise relationships can be extended to the prediction of small regulatory module networks by (1) inferring a Markov Blanket over a seed gene of interest and (2) constructing a BIC-penalized network. We constructed such regulatory module networks for well-studied mouse genes and found many intriguing examples of functionally coherent networks.

Brief Biography:

David Kulp is an assistant professor in Computer Science at the University of Massachusetts Amherst where his research is on understanding the effect of natural variation on biological regulatory processes, particularly the control of gene expression and splicing. Before coming to UMass, David was Vice President of Bioinformatics at Affymetrix, the major manufacturer of high density oligonucleotide microarrays, where he worked on genome annotation, microarray design, and algorithm development. David received his PhD in 2003 from UC Santa Cruz with David Haussler, where he developed novel semi-Markov HMMs for protein-coding gene prediction that were used to generate the gene sets in the initial sequencing and analysis of the fruit fly, mouse, and human genomes.