With the ever-increasing volumes of metagenomic sequencing data, the need for tools to infer the functions of microbial genes is growing as well. Using a combination of classical sequence comparison approaches, machine learning tools and modern AI methods, we are establishing computational workflows to predict the functions of microbial genes. We focus these efforts on poorly characterised organisms, for which we may only have metagenomic sequences, but no isolate, and on genes involved in various aspects of secondary metabolism including xenobiotics metabolism, biosynthetic pathways, complex carbohydrate metabolism and virulence factors. This project aims at generating novel hypotheses on how members of the human microbiome may shape community composition (through microbe-microbe interactions) or contribute to host health or disease processes, e.g. through pro-inflammatory or carcinogenic metabolites or through metabolization of medicinal drugs, environmental chemicals or dietary compounds that are associated with disease risks.