Ranking microbial metabolomic and genomic links in the NPLinker framework using complementary scoring functions.

Area of Science:

• Microbiology
• Metabolomics
• Genomics
• Natural Product Drug Discovery

Background:

• Microbial specialized metabolites are vital for biomedical applications, especially antibiotics.
• Linking microbial genomes to their metabolites (Biosynthetic Gene Clusters - BGCs) aids in discovering novel compounds.
• Current methods struggle to automatically and accurately link BGCs to metabolites due to biosynthetic knowledge gaps.

Purpose of the Study:

• To develop and present NPLinker, a software framework for linking genomic and metabolomic data.
• To improve the accuracy and efficiency of identifying true links between BGCs and metabolites.
• To address the bottleneck in natural product research caused by manual verification.

Main Methods:

• Utilizing paired omics data sets (genomic and metabolomic).
• Developing and applying multiple link-scoring functions, including a novel Input-Output Kernel Regression score.
• Standardizing and improving commonly used scoring metrics.
• Validating results using publicly available datasets with known links.

Main Results:

• Demonstrated that combining multiple link-scoring functions enhances the prioritization of true BGC-metabolite links.
• Introduced a novel, effective scoring method using Input-Output Kernel Regression.
• NPLinker successfully links genomic and metabolomic data, verified by existing validated datasets.

Conclusions:

• NPLinker provides an effective computational framework for linking microbial BGCs to specialized metabolites.
• The developed scoring functions significantly improve the accuracy of identifying novel natural products.
• This approach accelerates natural product research by automating and refining the discovery process.