Comparative Evaluation of Prediction Algorithms (CoEPrA) is a world-wide competition where teams make predictions of the chemical and biological properties of small molecules and peptides. This year, there were two types of competitions – Classification and Regression. Curt Breneman, Professor of Chemistry and Chemical Biology, led a team that placed first in the Regression competition. Winners were announced on July 31.

Rensselaer's team was an interdisciplinary group made up of professors, research scientists, graduate students, and undergraduate students. Group members included:

Kristin Bennett - Professor of Mathematical Sciences
Charles Bergeron - Graduate Student - Mathematics
Curt Breneman - Professor of Chemistry and Chemical Biology
Theresa Hepburn - Senior, Bioinformatics & Molecular Biology
Michael Krein - Senior, Chemistry and Chemical Biology
Min Li - Graduate Student, Information Technology
Steven Mulick - Senior, Chemistry and Chemical Biology
Sukumar Nagamani - Research Associate Professor of Chemistry and Chemical Biology (pending)
Matthew Sundling - Graduate Student, Chemistry and Chemical Biology

The contest organizer provides a dataset of molecules for which some experimental values are known, and then this information (along with any molecular descriptors the team may compute) is used to develop machine learning-based models to predict the same properties for a blind test set of molecules.

This is a classic quantitative structure activity relationship (QSAR) or quantitative structure-property relationships (QSPR) problem, and is quite pertinent to the drug-discovery and molecular optimization communities. The CoEPrA competition is similar to the CASP competition that pits one protein folding algorithm against another.

For more information, see the CoEPrA website at www.coepra.org.

Published August 8, 2006