Fig. 1.

Schematic overview of an algorithm for tRNA-derived cyclodipeptide identification and prediction. Given a microbial genome sequence as input, our algorithm uses a hidden Markov model to identify CDPSs and cluster them with surrounding biosynthetic and resistance enzymes, including tailoring enzymes specific to cyclodipeptide biosynthesis. CDPS active site residues are identified by multiple sequence alignment to a large database of CDPS sequences, and their aminoacyl-tRNA substrates are predicted by a naive Bayes classifier. Finally, tailoring reactions are executed to generate a predicted cyclodipeptide chemical structure