Genome-wide bioinformatic mining and evidence-based validation of PAL and CHS candidate genes in kasuri methi (Trigonella corniculata L.)

Chromosome-scale genomic resources are now available for Kasuri methi, however, reliable pathway annotation remains difficult because specialized-metabolism gene families are often paralogous and predicted gene models may be incomplete. We re-evaluated 117 pathway-oriented candidates from the 64,801 predicted protein-coding genes of Trigonella corniculata using an evidence-integrated workflow that combined curated Swiss-Prot homology, KO/EC concordance, protein-CDS-ORF consistency, Pfam profile-HMM searches, conserved-feature checks, family phylogeny, genomic context and targeted reconstruction. Seven phenylalanine ammonia-lyase (PAL) candidates contained PF00221 and retained the tested MIO-forming ASG tripeptide, while two chalcone synthase (CHS) candidates contained PF00195 and PF02797 across approximately 96% of their proteins and retained all four tested conserved functional residues. Two hydroxymethylglutaryl-CoA reductase (HMGR) candidates contained PF00368, and a 21-sequence, 377-aa conserved-domain phylogeny supported HMGR-family affinity; however, their 293-320-aa protein models were only about half the median length of full-length plant HMGR references and retained only partial or divergent diagnostic sequence features. LOX, oxidosqualene cyclase (OSC) and terpene synthase (TPS) candidates remained strongly truncated and were not rescued by homolog-guided reconstruction, while the tested NNMT and IMT/D-pinitol-associated models lacked sufficient full-length support for confident assignment. Overall, the analysis resolves the initial screen into nine strongly supported PAL/CHS family-level candidates, two partial HMGR-family models, and a set of unresolved or unsupported annotations that require further gene-model refinement rather than functional transfer from short local similarity alone.