RiPPMiner: a bioinformatics resource for deciphering chemical structures of RiPPs based on prediction of cleavage and cross-links

Abstract

Ribosomally synthesized and post-translationally modified peptides (RiPPs) constitute a rapidly growing class of natural products with diverse structures and bioactivities. We have developed RiPPMiner, a novel bioinformatics resource for deciphering chemical structures of RiPPs by genome mining. RiPPMiner derives its predictive power from machine learning based classifiers, trained using a well curated database of more than 500 experimentally characterized RiPPs. RiPPMiner uses Support Vector Machine to distinguish RiPP precursors from other small proteins and classify the precursors into 12 sub-classes of RiPPs. For classes like lanthipeptide, cyanobactin, lasso peptide and thiopeptide, RiPPMiner can predict leader cleavage site and complex cross-links between post-translationally modified residues starting from genome sequences. RiPPMiner can identify correct cross-link pattern in a core peptide from among a very large number of combinatorial possibilities. Benchmarking of prediction accuracy of RiPPMiner on a large lanthipeptide dataset indicated high sensitivity, specificity, accuracy and precision. RiPPMiner also provides interfaces for visualization of the chemical structure, downloading of simplified molecular-input line-entry system and searching for RiPPs having similar sequences or chemical structures. The backend database of RiPPMiner provides information about modification system, precursor sequence, leader and core sequence, modified residues, cross-links and gene cluster for more than 500 experimentally characterized RiPPs. RiPPMiner is available at http://www.nii.ac.in/rippminer.html.

Supporting supervised learning in fungal Biosynthetic Gene Cluster discovery: new benchmark datasets

“…Supervised learning was shown to perform well at BGC discovery in previous work that focused on handling bacteria data [5], [6]. Given that annotated data are needed to perform a supervised learning approach, we propose here fungal BGC datasets to support the development of this approach for fungi.…”

“…Supervised learning has been previously used to predicting bacterial BGCs [5], [6] and shown to perform well. Supervised learning methods however are developed primarily based on annotated datasets, for which all instances are labeled as belonging to a specific class.…”

Mining for Microbial Gems: Integrating Proteomics in the Postgenomic Natural Product Discovery Pipeline

“…With all the identified gene clusters, researchers have also created tools that link back from NPs to possible protein domain organization and harness this information to find the responsible BGCs from the databases . Genome‐mining algorithms like NRPSpredictor and RiPPMiner utilize machine‐learning technologies that form the new generation in bioinformatics. Harnessing the power of the artificial intelligence boom across every aspect of human life, these tools provide more information about predicted gene clusters and their likely product(s), whereby prediction accuracy is continuously improved.…”

Discovery and characterisation of an amidine-containing ribosomally-synthesised peptide that is widely distributed in nature

“…This is because the small precursor peptides that are ultimately transformed into the final product are often not annotated in genomes, and unlike with other natural product classes such as polyketides, terpenes and non-ribosomal peptides, the short biosynthetic pathways for RiPPs lack universally shared features. [7] Whilst some specific genome mining tools for RiPPs have been developed, [8][9][10][11] many of these tools rely on the identification of homology to known RiPP classes. Therefore, the opportunity to identify truly novel RiPP families, and subsequent untapped structural complexity, might be missed.…”