Motif Regressor: A Computational Method for Integrating Sequence Motif Discovery and Microarray Data

• Activity Number: 410
• Type: Contributed
• Date/Time: Thursday, August 15, 2002 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section*
• Author(s): Erin Conlon*+ and Xiaole Liu and Douglas Brutlag and Jun Liu
• Affiliation(s): Harvard University and Stanford University and Stanford University and Harvard University
• Keywords: Motif Discovery ; Sequence Data ; Microarray Data ; Correlation ; Transcription Control ; Computational Biology

Abstract

Despite a number of effective computational methods for DNA regulatory binding motif discovery that have been proposed and studied in the past decade, the integration of motif analysis and the genome-wide expression data has just begun. Available methods have limitations, and new tools are needed to better combine the sequence and expression information.

We have designed and implemented a novel method, Motif Regressor, to combine the advantages of existing methods and avoid some of their pitfalls. Our method uses a new deterministic motif-finding approach, MDscan, to find candidate motifs in the regulatory regions of genes most overexpressed (or underexpressed) in an experimental condition and refine the motifs using a subset of the remaining genes. Discovered motifs are confirmed with a stepwise regression to further utilize the microarray information. A final model includes motifs acting additively on expression. From one set of cell-cycle expression experiments, we found fourteen significant motifs classified in nine families, three of which were biologically known, two were predicted by other motif finding programs, and four are new putative motifs.