About PlantBGC
PlantBGC is a Transformer-based framework for plant BGC discovery.
It converts plant sequences into Pfam-domain representations,
predicts candidate BGC loci using a three-stage training strategy
(microbial supervision → label-free plant domain adaptation → optional weak supervision),
and returns downloadable results by email.
Output
Results will be sent to the provided email address with a download link.
Returned files may include:
*.bgc.tsv
Candidate BGC prediction table with locus coordinates and scores
*.pfam.tsv
Per-protein Pfam domain annotation and BGC-likeness scores
*.full.gbk
Full annotated GenBank file with all predicted features
*.bgc.gbk
GenBank file filtered to BGC candidate regions only
LOG.txt
Full run log for reproducibility and debugging
No online visualization is provided. Results are returned as downloadable files.
Example Usage
PlantBGC can also be run locally via the command line:
plantbgc Predict
--output
prediction_output/
--score
0.5
--min-proteins
3
input.fna
plantbgc Predict
--protein
--output
prediction_output/
proteins.fasta
Documentation
Full documentation including installation instructions, data format specifications,
and parameter descriptions are available in the
PlantBGC GitHub repository.