How to Use antiSMASH for Biosynthetic Gene Cluster Analysis

Closing the gap between metabolite identification and biological understanding requires integrating genomic and metabolomic data. Genome mining tools like antiSMASH identify biosynthetic gene clusters encoding secondary metabolite production. By comparing genetic signatures across organisms and matching them to metabolomic profiles, researchers link molecular structures to their biological producers. This integration enables identification of which microbes produce which compounds, accelerating natural product discovery. The ultimate goal is creating systems where new metabolomics data can automatically retrieve existing annotations, dramatically reducing the annotation burden and enabling more efficient exploration of natural chemical diversity.

Dereplication accelerates natural products research by quickly identifying known compounds, allowing focus on novel leads. Strategies include: targeted analysis with database searching, bioassay-guided fractionation analyzing only active fractions, analyzing all fractions with database searches, and metabolomics-based approaches combining parallel bioactivity assays with chemometric analysis. Molecular networking represents a powerful computational approach requiring high-resolution accurate mass MS/MS data. Data conversion to mzML format enables upload to platforms like GNPS (Global Natural Products Social Molecular Networking). The algorithm generates molecular networks based on fragmentation pattern similarities—compounds sharing structural features cluster together. Known compounds in databases enable rapid identification; novel compounds can be structurally elucidated by comparing networks to known compounds. This approach enables high-throughput profiling of natural extracts and rapid identification of bioactive compounds based on accurate mass and fragmentation data. Case studies demonstrate successful identification of antitrypanosomal compounds from plant extracts using this workflow.

This section explains how combining genomic and metabolomic data accelerates discovery of natural products. Genome mining identifies biosynthetic gene clusters encoding specialized metabolites, while metabolomics provides complementary stereochemical information encoded in genes but difficult to obtain from mass spectrometry alone. The Paired Omics Data Platform records links between genome sequences and LC-MS/MS datasets, enabling strain-based correlation matching and feature-based matching. This integration reduces false positives by requiring consistent co-occurrence of gene clusters and metabolites across biological replicates. Combined with chemical class ontology prediction, this approach accelerates discovery of plant and microbial natural products relevant to agriculture and medicine. The IOMICA project exemplifies this integrated approach, demonstrating how computational frameworks can bridge genetic potential with actual metabolic outputs.

Dereplication is a major bottleneck in metabolomics research. Taxonomy-based approaches use sample classification knowledge to prioritize candidate structures commonly found in that family, narrowing the search space. High-resolution mass matching compares measured molecular weights against database entries with 5 ppm tolerance, requiring HRMS and tandem capabilities. MS/MS spectral matching uses global databases like GNPS, NIST, and MBank. When experimental data is insufficient, in silico approaches using Sirius and CFMID predict structures by comparing MS2 data against database entries. This requires persistence as dereplication is described as passing through a 'needle in a haystack' challenge.

The Natural Products Atlas is an open-access database containing over 25,000 microbial natural products that can be integrated with untargeted metabolomics workflows to accelerate natural product discovery by providing chemical diversity visualization, taxonomic filtering, and cluster-based identification strategies that improve compound annotation accuracy.

This section covers the gene identification and validation process. Researchers identified 134 UGT candidates and 34 SCPL candidates through BLAST analysis, then selected three of each based on differential expression patterns. Heterologous expression in Nicotiana benthamiana with GFP markers allowed testing of gene combinations. One combination (UGT + SCPL) successfully produced clovamide, confirming the pathway. Additionally, ICS-1 has a frameshift mutation in the UGT gene causing premature truncation, explaining its inability to produce clovamide. This demonstrates how genetic variation in key enzymes can determine metabolic capabilities.

The complete gene knockout workflow involves multiple interconnected steps: (1) designing a construct with homologous sequences, (2) introducing the construct into host cells, (3) allowing homologous recombination at the target locus, (4) applying dual selection (positive and negative) to isolate correctly targeted cells, and (5) verifying the knockout through molecular techniques like PCR and Southern blotting. Each step requires careful optimization to achieve successful gene disruption.

To verify a reconstructed pathway, test each mutant with compounds at different positions: (1) For mutant 1 (step 1 mutation), growth only occurs with D (end product); (2) For mutant 2 (step 2 mutation), growth occurs with A, B, and D but not C; (3) For mutant 3 (step 3 mutation), growth occurs with A, B, and D but not B and C. This systematic testing confirms the pathway order and validates the gene dissection results.

To confirm that a candidate enzyme has the predicted function, experimental validation is necessary. This can be done by expressing the candidate gene in a heterologous system (like bacteria) and testing whether the enzyme produces the expected product. Alternatively, researchers can use genetic approaches like gene knockdown or knockout to see if the expected biosynthetic product is reduced or eliminated. Only through such experiments can the predicted function be confirmed.

Characterizing novel natural product pathways presents unique challenges when heterologous hosts lack necessary tailoring enzymes. Cyanobacterium, first chlorinated natural product from freshwater (1982), has phytoxic activity inhibiting photosynthetic electron transport. Initial heterologous expression produced precursor-like molecules lacking final modifications, revealing that E. coli lacked required enzymes. Carbon labeling with tyrosine and valine identified incorporation patterns, establishing building blocks. Systematic knockout studies identified essential genes (C, E, F) by observing complete loss of product formation. The final ring-forming enzyme proved novel and was renamed furanolide synthase. Symbiotic cyanobacteria form partnerships with diverse hosts including lichens, plants (cycads, moss), marine dinoflagellates, and animals, creating unique metabolic environments likely producing specialized natural products absent in free-living cyanobacteria. However, genomic data remains extremely limited—NCBI contains less than 1% of all cyanobacterial genomes, with only 7 lichen symbiont genomes (all Nostoc). Collaborative initiatives now aim to address this gap through large-scale sequencing targeting diverse symbiotic cyanobacteria from Australia, Germany, South Africa, and other regions, providing high-quality genomes enabling research on bioactive natural products.

Natural products are bioactive compounds from microbes, fungi, and plants serving ecological functions like defense and competition. Secondary metabolites, not essential for survival but providing evolutionary benefits, are the source of drugs like penicillin and rapamycin. The golden age of antibiotic discovery (1940s-1970s) ended around 1990s, with no new antibiotics since, compounded by the antibiotic resistance crisis causing an estimated 2 million deaths annually by 2050. Traditional methods are laborious and suffer from rediscovery of known compounds. Genome mining addresses this by computationally scanning genomes to identify biosynthetic gene clusters (BGCs)—groups of collocated genes encoding enzymes for natural product synthesis. However, 98-99% of microbial species cannot be cultivated (microbial dark matter), meaning traditional methods access only 1% of potential diversity. Metagenomics solves this by directly sequencing environmental DNA, capturing genomic information from uncultured organisms. The BGC Atlas addresses limitations of existing databases by providing a centralized resource covering nearly all publicly available metagenome data, metadata-aware for correlating samples with environmental information, searchable for sample comparison, and explorable via web interface. Analysis of 35,000 metagenomic assemblies reveals a completely unexplored chemical space with almost no similarity to known compounds. Biosynthetic gene clusters show strong habitat specificity, with most GCFs restricted to specific environmental variables. Current limitations include data source bias, incomplete BGCs from short Illumina contigs, and lack of taxonomic information. Future improvements include integrating data from JGI IMG/M, incorporating machine learning-based tools like BiG-SCAPE, adding functional bioactivity predictions, and expanding to long-read nanopore sequencing. Ultra-deep nanopore sequencing revealed less than 0.6% of detectable species had been captured previously, with over 13,000 BGCs detected and matches for only 108, demonstrating less than 1% of environmental chemical diversity has been discovered.

This video presents an integrated computational pipeline that combines multiple algorithms (anti-SMASH and DBGC) to detect biosynthetic gene clusters (BGCs) and ribosomally synthesized and post-translationally modified peptides (RiPPs) in microbial metagenomes, demonstrating how combining different detection methods provides more comprehensive insights into specialized metabolite biosynthesis across diverse ecosystems.

Researchers have applied Gecko to mine massive datasets including 34,000 ocean metagenomes, nearly a million Progenomes 3 genomes, and 2.5 million bacterial genomes. All predicted BGCs are made publicly available. To identify novel antibiotics among millions of BGCs, researchers use clustering algorithms to create gene cluster families (GCFs), assigning predicted function based on similarity to experimentally validated BGCs. Analysis of 2.8 million BGCs revealed that the vast majority of gene cluster families contain no experimentally validated BGCs, indicating vast unexplored biosynthetic potential. While genomic data is valuable, effective AMR research requires interdisciplinary collaboration across microbiologists, clinicians, epidemiologists, bioinformaticians, and social scientists to fully leverage its potential.

Genomic mining searches databases to identify genes of interest for bioprospecting. Two approaches exist: targeted mining for specific organisms and comparative genomics analyzing multiple genomes. The bioinformatics pipeline includes: sequencing (short-read or long-read), assembly, gene prediction, functional annotation, and biosynthetic gene cluster (BGC) prediction. Public databases like NCBI provide access to genomic data. BGCs include NRPS, PKS, and hybrid clusters that produce secondary metabolites. This approach has been used to discover novel enzymes, antibiotics, and other valuable compounds from environmental bacteria.

Metagenomics is the genomic analysis of microbial communities by directly extracting and analyzing DNA from environmental samples, enabling the study of microorganisms in their natural habitats without requiring pure culture cultivation; this approach employs two main methodologies—functional metagenomics (constructing libraries to screen for specific enzymatic activities) and metagenomic DNA sequencing (using 16S rRNA gene amplification for phylogenetic analysis or whole-genome assembly for comprehensive functional and taxonomic annotation)—to reveal microbial diversity, evolutionary relationships, and ecological functions within complex communities.