HTS stages

Figure 3. Retrospective analysis of HTS data.

the step where a properly trained neural network should make similar decisions.

Figure 3 illustrates the results. Whereas the percentage of drug-like compounds with a score >0.3 remains rather constantly on the 50-60% level for the first three HTS stages (screening, after primary screen (% inhibition), and after secondary screen (IC50)), it jumps up to 70% for the enzyme assays and to nearly 100% for the receptor assays after manual inspection by medicinal chemists (whether the difference between receptors and enzymes is systematically or not should be discussed when more data are available).

Two conclusions can be drawn from this result. First, there is no direct correlation between general drug-likeness and activity in certain biological assays - also non-drug-like compounds can cause signals. Secondly, the trained neural network for the drug-likeness prediction behaves similar to trained medicinal chemists and tends to select the same sort of compounds.

Therefore, efforts for screening can be cut significantly by applying this filter before HTS.

0 0

Post a comment