PlantBGC: Transformer for Plant BGC Discovery via Label-Free Domain Adaptation and Weak Supervision

2026-07-28

AI Summary

This study addresses the challenge of accurately identifying plant biosynthetic gene clusters (BGCs) at a genome-wide scale, hindered by scarce labeled data and the high false-positive rates of rule-based methods across domains. The authors propose a novel approach that combines an encoder-only Transformer with unsupervised domain adaptation, modeling gene clusters as ordered sequences of Pfam domains. Pretrained on microbial BGCs via masked language modeling, the model is transferred to plant genomes and further refined using Gene Ontology (GO) and KEGG functional annotations as weak supervision signals. Evaluated on 34 known plant BGC loci, the method increases the 100% coverage recall from 29.4% to 67.6%, reduces the median predicted region length to 27.8% of that produced by plantiSMASH, and decreases the proportion of primary metabolism-like false positives by over 45%.

Abstract

Plant biosynthetic gene clusters (BGCs) encode specialized-metabolite pathways, yet curated plant BGC labels remain scarce, hindering supervised discovery at genome scale. Existing plant BGC mining tools are largely signature- and rule-driven and do not fully leverage recent advances in contextual representation learning for modeling long-range domain context and controlling false positives under strong domain shift. We seek an AI-assisted workflow that narrows experimental search space by transferring supervision from well-annotated microbial BGCs to plant genomes. We present PlantBGC, representing genomes as ordered Pfam-domain sequences and learning BGC-likeness with an encoder-only Transformer trained on MIBiG microbial BGCs and adapted to plants via label-free masked language modeling. On microbial benchmarks, PlantBGC achieves token-level AUC = 0.988 (10-fold CV) and 0.979 (leave-class-out). On plants, adaptation improves known-BGC recovery on n = 34 curated loci under strict 100% coverage, increasing recovery from 29.4% to 67.6% and indicating more complete boundaries. GO/KEGG-derived weak supervision reduces proxy primary-like ratio by 48.40% (GO) and 45.20% (KEGG), with consistent per-species reductions (paired Wilcoxon p = 1.53e-5). Compared to plantiSMASH, PlantBGC yields more compact loci on matched regions (median length ratio = 0.278; 93.8% of pairs are shorter).

Problem

• plant biosynthetic gene clusters
• label scarcity
• domain shift
• supervised discovery
• genome-scale mining

Innovation

• Transformer
• domain adaptation
• weak supervision
• biosynthetic gene clusters
• label-free learning