Abstract Details

Tracking Compound and Project Progress Through Multi-objective-based Compound Quality Assessment

J. Willem M. Nissink1
1AstraZeneca (Alderley Park)
Quality-scoring methods designed to evaluate multiple objectives concurrently are useful tools to analyse output from drug design efforts and rationalize selection for progression. We describe a systematic approach to prioritizing molecules based on emerging experimental data in the course of a drug-discovery project.

The compound-quality framework we propose is based on experimental data and reflects a falsification approach, where unfavourable data is used to penalize a quality measure. We use simple desirability functions and introduce novel aggregated and multi-parameter desirability functions to reflect commonly used ‘rules’.

Quality scores based on such desirabilities are generated using a collation metric commonly used in multi-parameter optimization, as well as a more lenient distance-based metric. The quality score we propose deals transparently with missing data, a key requirement in drug-hunting projects where data is often limited and generated depending on a perception of quality which is often project-specific.

We further outline a way to estimate confidence in the interpretation of such a compound-quality measure, given the often limited or imperfect data for the compound at hand. In combination, scores and associated confidences provide systematic insight in the quality of emerging chemical equity, and we illustrate this with a few examples.

The approach also allows us to track the quality of lead compounds over time, yielding insight into the progress of drug-hunting teams. Applied appropriately, use of such scores may have further application in risk and resource management of a drug portfolio, when used across projects and calibrated for each project.

Return to Programme