The goal of drug discovery is to find potent and selective inhibitors for particular targets. A good ligand thus should be small (for better bio-availability) and potent.

The larger is the inhibitor, the more there are ways to interact with specific features of a certain binding site. On the other hand a larger ligand may be flexible enough to “find” a place to bind to another protein. If the line of thought is correct, then small ligands are promiscuous, larger ligands are more selective.

To check the argument we analyzed the available experimental binding data from BindingDB, ChemBL and NIH assays. First we extracted all the ligands and then clustered the molecules using standard Daylight fingerprints from OpenBabel (using Tanimoto similarity cutoff 0.85). Then we used the assays data to collect all the activities associated with the ligands clusters. Finally we plotted the average molecular weight of the cluster members vs. the number of assays (targets) associated with the cluster.

The outlook is quite unexpected. The promiscuity (the number of targets associated with the molecules in the cluster, the horizontal axis) of a ligand does not seem to depend on the molecular weight (the vertical axis). The decrease of the molecular weight on the right of the graph comes from the lack of data in the dataset (that is why the noise).

The latest graph should be also compared with the data on the activity of drugs vs. their toxicity. The promiscuity of a molecule is a signature of side effects and possible toxicity. Neither the size or the activity of a molecule does not appear to correlate with its promiscuity (or toxicity).

Related posts:

1. How drug-like are vendor libraries?
2. How good are biological data – II: Trombine, GSK, GPCR
3. High Throughput Screening of Transcription factor DLX5
4. Protein Flexibility and False Positives detection.
5. Docking validation study: classic example, thrombine