From DNA sequences to chemical structures--methods for mining microbial genomic and metagenomic data sets for new natural products

Rapid mining of large genomic and metagenomic data sets for modular polyketide synthases, non-ribosomal peptide synthetases and hybrid polyketide synthase/non-ribosomal peptide synthetase biosynthetic gene clusters has been achieved using the generic computer program packages ClustScan and CompGen. These program packages perform the annotation with the hierarchical structuring into polypeptides, modules and domains, as well as storage and graphical presentations of the data. This aims to achieve the most accurate predictions of the activities and specificities of catalytically active domains that can be made with present knowledge, leading to a prediction of the most likely chemical structures produced by these enzymes. The program packages also allow generation of novel clusters by homologous recombination of the annotated genes in silico. ClustScan and CompGen were used to construct a custom database of known compounds (CSDB) and of predicted entirely novel recombinant products (r-CSDB) that can be used for in silico screening with computer aided drug design technology. The use of these programs has been exemplified by analysing genomic sequences from terrestrial prokaryotes and eukaryotic microorganisms, a marine metagenomic data set and a newly discovered example of a 'shared metabolic pathway' in marine-microbial endosymbiosis. Key words: polyketides, non-ribosomal peptides, Actinobacteria, homologous recombination