SeMPI

SeMPI predicts and identifies type I polyketides and nonribosomal peptides from genomic biosynthetic gene clusters to link gene clusters with known natural products or assess novelty.

Key Features:

• Polyketide and NRPS prediction: Predicts scaffolds of type I polyketides and nonribosomal peptides by leveraging their modular assembly from biosynthetic building blocks.
• Biosynthetic gene cluster analysis: Performs genome mining to detect and evaluate biosynthetic gene clusters responsible for secondary metabolite biosynthesis.
• Refined prediction pipeline: Implements a refined pipeline for scaffold generation and high-quality prediction of secondary metabolite structures.
• Robust screening algorithm: Detects homologous structures even in partial or incomplete biosynthetic gene clusters to improve identification sensitivity.
• Benchmarking: Includes a cluster detection algorithm benchmarked against antiSMASH v5 using an evaluation of 559 gene clusters.
• Novelty assessment: Enables linking of gene clusters to known natural products or estimation of novelty for uncharacterized clusters.
• Scalability for genomic data: Designed to handle large volumes of published genomic data for comprehensive genome mining.

Scientific Applications:

• Natural product discovery: Identifies candidate biosynthetic gene clusters encoding type I polyketides and nonribosomal peptides for downstream natural product characterization.
• Biosynthetic pathway elucidation: Facilitates reconstruction of modular NRPS and PKS scaffolds to support biosynthetic pathway annotation and hypothesis generation.
• Comparative cluster analysis: Enables comparison of gene clusters against known structures to detect homologs and assess structural relatedness.
• Novelty and dereplication: Supports assessment of cluster novelty and dereplication by linking predicted scaffolds to known natural products.

Methodology:

Refined prediction pipeline that predicts scaffolds by leveraging modular assembly of NRPS/PKS; a robust screening algorithm detects homologous structures in partial or incomplete biosynthetic gene clusters; cluster detection algorithm was benchmarked against antiSMASH v5 using 559 gene clusters.