NRPSpredictor2--a web server for predicting NRPS adenylation domain specificity.

Area of Science:

• Biochemistry
• Computational Biology
• Genomics

Background:

• Non-ribosomal peptide synthetases (NRPS) produce vital secondary metabolites, including antibiotics like vancomycin.
• Identifying new NRPS- A-domains via genome sequencing is key for discovering novel bioactive compounds.
• Predicting A-domain substrate specificity aids in annotating these gene clusters.

Purpose of the Study:

• To develop an improved computational tool for predicting NRPS A-domain substrate specificity.
• To enhance the identification and annotation of gene clusters responsible for secondary metabolite production.
• To provide a reliable predictor for both bacterial and fungal A-domains.

Main Methods:

• Utilized Support Vector Machines (SVM) for predicting A-domain specificity on four hierarchical levels.
• Developed a predictor based on sequence signatures and machine learning, improving upon previous work (NRPSpredictor).
• Modeled the predictor's applicability domain to assess reliability for new A-domains.

Main Results:

• Achieved high prediction accuracy (F-measure > 0.89) for three general specificity levels and 0.80 for the most detailed level.
• Successfully developed a predictor for fungal A-domains with an F-measure of 0.84 for gross physicochemical properties.
• The predictor is available as a web service for public use.

Conclusions:

• The improved predictor significantly enhances the ability to determine NRPS A-domain substrate specificity.
• This tool facilitates the discovery and annotation of novel secondary metabolites, particularly antibiotics.
• The predictor offers a valuable resource for researchers in natural product chemistry and drug discovery.