Summary
antiSMASH ships 572 custom HMM profiles for biosynthetic gene cluster (BGC) detection. These are standard HMMER3 format and work perfectly with Aksha's --hmm_in pathway. We should add them as a first-class installable database (like PFAM, KOFAM, DefenseFinder, etc.) so users can run --installed_hmms antismash instead of manually extracting them.
Source URLs
All profiles are checked into the antiSMASH GitHub repo (MIT license):
Tested against commit: antismash/antismash@master (2026-03-14 snapshot).
Integration notes
-
HMMER format mismatch: The detection HMMs are HMMER3/f (2015+) but the NRPS/PKS domain files are HMMER3/b (2010-2013). hmmconvert must be run on the older files before concatenation, or pyhmmer will fail mid-parse with ValueError: Invalid format in file.
-
GA thresholds: 208/572 profiles have GA thresholds. Recommend --cascade as default (use GA where available, fall back to e-value).
-
Cluster rules: antiSMASH also ships cluster detection rules (strict.txt, relaxed.txt, loose.txt) that define which combinations of HMM hits constitute valid BGCs. These are outside Aksha's scope but could be referenced in output documentation.
-
Download script: Should clone with --depth 1 --sparse, checkout only the three data directories, run hmmconvert on the NRPS/PKS files, concatenate, and hmmpress.
Validation
Tested on DPANN dataset (3,564 genomes, 3.4M proteins):
--cascade: 20,903 hits, 10/572 profiles with hits--evalue 1e-10: pending (expect more hits from profiles lacking GA thresholds)
Most hits are generic enzyme families (Aminotran, AMP-binding). True biosynthetic-specific hits: 4 Condensation domains, 19 chalcone/stilbene synthase, 8 autoinducer synthase — consistent with DPANN archaea having minimal secondary metabolism.
Summary
antiSMASH ships 572 custom HMM profiles for biosynthetic gene cluster (BGC) detection. These are standard HMMER3 format and work perfectly with Aksha's
--hmm_inpathway. We should add them as a first-class installable database (like PFAM, KOFAM, DefenseFinder, etc.) so users can run--installed_hmms antismashinstead of manually extracting them.Source URLs
All profiles are checked into the antiSMASH GitHub repo (MIT license):
antismash/detection/hmm_detection/data/antismash/detection/nrps_pks_domains/data/nrpspksdomains.hmm(60),abmotifs.hmm(99),ksdomains.hmm(5),dockingdomains.hmm(2)antismash/modules/nrps_pks/data/aa-activating.aroundLys.hmm)Tested against commit:
antismash/antismash@master(2026-03-14 snapshot).Integration notes
HMMER format mismatch: The detection HMMs are HMMER3/f (2015+) but the NRPS/PKS domain files are HMMER3/b (2010-2013).
hmmconvertmust be run on the older files before concatenation, or pyhmmer will fail mid-parse withValueError: Invalid format in file.GA thresholds: 208/572 profiles have GA thresholds. Recommend
--cascadeas default (use GA where available, fall back to e-value).Cluster rules: antiSMASH also ships cluster detection rules (
strict.txt,relaxed.txt,loose.txt) that define which combinations of HMM hits constitute valid BGCs. These are outside Aksha's scope but could be referenced in output documentation.Download script: Should clone with
--depth 1 --sparse, checkout only the three data directories, runhmmconverton the NRPS/PKS files, concatenate, andhmmpress.Validation
Tested on DPANN dataset (3,564 genomes, 3.4M proteins):
--cascade: 20,903 hits, 10/572 profiles with hits--evalue 1e-10: pending (expect more hits from profiles lacking GA thresholds)Most hits are generic enzyme families (Aminotran, AMP-binding). True biosynthetic-specific hits: 4 Condensation domains, 19 chalcone/stilbene synthase, 8 autoinducer synthase — consistent with DPANN archaea having minimal secondary metabolism.