Novel Antibiotics from a ‘White Box’ 2D Structural Fingerprint Decision Tree

The paper is concerned with repurposing drugs based on chemical similarity to existing drugs, with an application to antibiotics. A simple ‘white box’ 2D chemical fingerprint-based decision tree approach is shown to largely recapitulate a neural network study in the literature. In particular, the repurposing of halicin is shown to be based on an explicit fingerprint pattern, unlike the neural network ‘black box’ methodology.