A new in silico tool for predicting drugs’ risk of metabolism and bioaccumulation by gut microbiota

Microbial metabolism and bioaccumulation

Many drugs are now known to be metabolised and/or bioaccumulated by gut bacteria. The importance of these drug-bacteria interactions have been demonstrated by a number of human studies, which have shown the significant impact of microbial metabolism on patients’ pharmacokinetics. For example, patients’ dose requirements of two critical drugs, tacrolimus and digoxin, have been correlated with the abundance of drug-metabolising bacteria in their guts.

Knowledge of microbial transformation and bioaccumulation in the drug development phase could provide valuable insight into investigational drugs’ pharmacokinetic, and even toxicity, profiles. Through understanding how drug candidates interact with the microbiome, developers could identify which candidates may show pharmacokinetic variability, or be chemically transformed to toxic metabolites, in patients. Such knowledge could inform the progression of drug candidates from the drug discovery to development phase, thus allowing developers to optimise the success rate of their pipelines. However, experimentally screening hundreds of drug candidates’ susceptibility to microbial metabolism or bioaccumulation is often not feasible for drug developers, due to the time and resources required.

Machine learning accelerates the prediction of microbiome interactions

In a recent study, we developed a machine learning model to predict drugs’ susceptibility to metabolised and bioaccumulated by gut bacteria. Using 455 drug-microbiota interactions from the literature, an extremely randomised trees classifier was built that could predict drugs’ fate with over 80% precision. With this model, drug developers can input the chemical fingerprint of their investigational drugs and achieve high throughput predictions of their microbiome interactions in seconds to minutes.

The full article describing our model, in addition to the code required to use it, is available to access for free here: https://www.mdpi.com/1999-4923/13/12/2001.