Medicinal chemistry tools: making sense of HTS data

The main problem in QSAR modeling from the high throughput screening (HTS) data is that by definition, it produces only a small proportion of hits against a given assay. This leads to a very small statistical significance of the hits in comparison with the "noise". Analysis based purely on the "hit" compounds removes useful information about the biological response of all the test compounds. What is needed is an analysis technique that increases the significance of the active compounds, while using the information present in the original data. In this paper we present a method for application of intelligent filtering of the data to improve statistical significance of the active compounds to generate predictive models that provide medicinal chemists with a powerful tool for both optimizing compounds and mining screening candidates in libraries.