Machine Learning for Rational Drug Design

From 1991-1993, I worked as a Senior Scientist at Arris Pharmaceutical Corporation in South San Francisco, CA. One of the central goals of the research team at Arris is to apply machine learning (and other advanced computational techniques) to improve the drug design process. During the drug development process, it is common for chemists to synthesize and test many molecules with the goal of finding a molecule that binds tightly to a desired binding site. We developed several machine learning algorithms that can analyze such binding data and construct hypotheses about the requirements for tight binding. These hypotheses can be applied to guide the design of new molecules and to prioritize the synthesis of those molecules. These algorithms are currently being applied at Arris to several drug development projects.

• Dietterich, T. G., Lathrop, R. H., Lozano-Perez, T. (1997) Solving the multiple-instance problem with axis-parallel rectangles. Artificial Intelligencem, 89(1-2), 31-71. Postscript preprint (The last 3 figures are not available online).

• Jain, A. N., Dietterich, T. G., Lathrop, R. H., Chapman, D., Critchlow, R. E., Bauer, B. E., Webster, T. A., Lozano-Perez, T. (1994). Compass: A shape-based machine learning tool for drug design. Computer-Aided Molecular Design, 8 (6) 635--652.

• Dietterich, T. G., Jain, A., Lathrop, R., Lozano-Perez, T. (1994). A comparison of dynamic reposing and tangent distance for drug activity prediction. Advances in Neural Information Processing Systems, 6. San Mateo, CA: Morgan Kaufmann. 216-223. Postscript preprint.