A probabilistic approach to high throughput drug discovery

Area of Science:

• * Computational chemistry and cheminformatics.
• * Drug discovery and medicinal chemistry.
• * Quantitative structure-activity relationship (QSAR) modeling.

Background:

• * High throughput screening (HTS) generates large datasets for drug discovery.
• * Traditional QSAR models often rely on regression, which may not suit binary activity data.
• * Efficiently selecting building blocks for virtual combinatorial libraries is crucial.

Purpose of the Study:

• * To present a new methodology for constructing probabilistic QSAR models from HTS data.
• * To utilize these models for selecting building blocks for virtual combinatorial libraries.
• * To demonstrate the methodology's applicability in designing focused libraries for specific targets.

Main Methods:

• * Development of a probabilistic QSAR model based on statistical probability estimation, not regression.
• * Application of the model to select substituents for virtual combinatorial libraries.
• * Construction of focused libraries targeting cyclic GMP phosphodiesterase type V and acyl-CoA:cholesterol O-acyltransferase.

Main Results:

• * The probabilistic QSAR methodology successfully identified effective building blocks.
• * The approach enabled the selection of active compounds using only binary (pass/fail) activity data.
• * Two focused virtual combinatorial libraries were constructed, demonstrating the method's utility.

Conclusions:

• * The presented methodology offers a robust alternative to regression-based QSAR for binary data.
• * This approach facilitates the efficient design of targeted compound libraries.
• * The probabilistic QSAR model shows promise for accelerating drug discovery by optimizing library design.