Leveraging Chemical Insights in Computational Identification of Interesting Biosynthetic Gene Clusters

Natural products have been shown to have diverse applications in fields such as agriculture and medicine. The diminishing returns from screens of traditional chemical libraries coupled with the reduced costs of genome sequencing have fuelled a renewed interest in mining genetic sequences for biosynthetic gene clusters (BGCs). This has led to the development of methods to analyze genome sequences to identify potential BGCs. These have been successful at identifying gene clusters, particularly those which contain polyketide synthase or non-ribosomal peptide synthetase modules. However, the success of these algorithms in finding potential gene clusters has necessitated the need for further refinement and classification: the amount of putative gene clusters exceeds our ability to experimentally characterize them. While advances in DNA synthesis, cloning, and heterologous expression will increase our ability to characterize a potential gene cluster, methods to select which clusters to prioritize are still needed.

We present a framework that does this by examining the mechanistic steps involved in the biosynthesis of specific chemical moieties. By focusing on the chemical steps and the corresponding enzymes required to produce a specific structure, we can quickly screen in silico large genomic databases to a small, experimentally tractable list of potential BGCs that are likely to produce the targeted chemical moiety. Furthermore, by focusing on the relatively smaller space of biologically available starting products and chemical reactions, this framework can uncover novel, uncharacterized classes of BGCs complementing the discovery and annotation provided by existing methods.