Abstract
Significant progress has been made in the past few years on the computational identification of biosynthetic gene clusters (BGCs) that encode ribosomally synthesized and post-translationally modified peptides (RiPPs). This is done by identifying both RiPP tailoring enzymes (RTEs) and RiPP precursor peptides (PPs). However, identification of PPs, particularly for novel RiPP classes remains challenging. To address this, machine learning has been used to accurately identify PP sequences. Current machine learning tools have limitations, since they are specific to the RiPPclass they are trained for and are context-dependent, requiring information about the surrounding genetic environment of the putative PP sequences. NeuRiPP overcomes these limitations. It does this by leveraging the rich data set of high-confidence putative PP sequences from existing programs, along with experimentally verified PPs from RiPP databases. NeuRiPP uses neural network archictectures that are suitable for peptide classification with weights trained on PP datasets. It is able to identify known PP sequences, and sequences that are likely PPs. When tested on existing RiPP BGC datasets, NeuRiPP was able to identify PP sequences in significantly more putative RiPP clusters than current tools while maintaining the same HMM hit accuracy. Finally, NeuRiPP was able to successfully identify PP sequences from novel RiPP classes that were recently characterized experimentally, highlighting its utility in complementing existing bioinformatics tools.
Highlights
Specialized metabolites from bacteria have been a source of bioactive chemical compounds with myriad applications especially in the pharmaceutical and agrochemical industries[1]
In order to check that the high accuracy was not due to the neural network being overfit to the data, the models were trained on a dataset that randomly excluded 15% of the positive dataset (550 sequences), and 8.6% of the negative set (1650 sequences)
Peptide sequences classified as NeuRiPP hits show a similar or higher hidden Markov models (HMMs) hit rate to precursor peptide predictions in existing tools
Summary
Specialized metabolites from bacteria have been a source of bioactive chemical compounds with myriad applications especially in the pharmaceutical and agrochemical industries[1]. Proper identification of PPs is an important aspect of in silico RiPP BGC analysis as knowledge of the PP sequence can aid in structure elucidation and provide information on the molecular interactions between the RTEs and the PP5 To this end, several methods have been developed to identify putative PPs in regions in proximity to RTEs. To this end, several methods have been developed to identify putative PPs in regions in proximity to RTEs This typically involves a two-step process where sequences to be screened are first identified either through the use of gene-finding software[5,7], or from identifying open reading frames (ORFs) of specified length in the proximity of RTEs6,8. Prodigal-short was used to find putative PPs in proximity to RTEs, and peptide similarity network analysis of the identified PPs was used to identify new RiPP classes[5] This demonstrated the potential of using gene-finding software as a starting point for identifying novel RiPPs; the number of likely coding sequences from this approach was still large.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
