Statistical Analysis of Large Scale Screening Data

Combinatorial chemistry, a means for rapidly synthesizing large libraries of molecules, is an activwith the promise of dramatically accelerating and improving the process of drug discovery. The screelibraries is expected to help medicinal chemists find drug candidates quickly, but exactly what divepurpose remains controversial and somewhat vague. The principal objective of this project is to placmathematically rigorous footing by constructing a space of chemical descriptors such that biologicalpharmacologically relevant assays correlates with regions of that space. Overall goals include 1) thmetric for ranking libraries of molecules, 2) the precise assessment of the relative information provalued biological measurements, and 3) the development of an improved statistical inference procedurhypothesis generation. Phase I has several specific goals: 1) the construction of a metric space ofdescriptors; 2) the assessment of the descriptors using a novel, information-theoretic characterizatsignificance of pharmacophores generated using these descriptors; 3) a statistical characterizationbetween regions in descriptor space and pharmacological activity.