Druglikeness score


A method for classifying compounds into drugs and non-drugs was developed recently [4]. The intention of such an approach was to model the intuition

neural network

Figure I. Neural network approach for classifying compounds.

and experience of medicinal chemists beyond certain advantageous or disadvantageous (e.g., reactive) substructures. This is somehow a complex feeling for suitability, synthesizability, stability, bioavailability, or toxicity. The drug-likeness approach is based on a very simple type of fingerprint and a neural network. Figure 1 illustrates this schematically. A given chemical compound is translated into a suitable descriptor that is forwarded to a trained neural network which in turn gives back an estimate of a certain compound class the molecule belongs to (e.g., drug or non-drug).

The main idea behind this approach is that the same knowledge which medicinal chemists achieve in their career is in the form of many examples

0 0

Post a comment