Anti-MRSA bacteriocin models checked with Ramachandran plots
Researchers in Thailand used ProteinIQ's Ramachandran plot tool to validate AlphaFold3 models of bacteriocin candidates from anti-MRSA palm sugar bacteria.
- Authors
- Chonticha Romyasamit, Komwit Surachat, Nawanwat C. Pattaranggoon, Morteza Saki, et al.
- Published in
- Antibiotics · 2026-10 · Walailak University
- ProteinIQ in this study
- AlphaFold3 models of three bacteriocin candidates were checked for backbone geometry with ProteinIQ's Ramachandran plot tool.
- Tools used
- Ramachandran Plot
The research
Methicillin-resistant Staphylococcus aureus (MRSA) causes skin infections, pneumonia, bloodstream infections, and device-associated infections, and its resistance to most β-lactam antibiotics makes it hard to treat. Bacteriocins, the antimicrobial peptides made by lactic acid bacteria, are one source of alternatives that researchers keep returning to.
Earlier work by the same group showed that two Lacticaseibacillus paracasei strains isolated from Palmyra palm sugar in southern Thailand, WU0904 and WU2302, inhibit MRSA and reduce its biofilm formation. What the strains did not reveal was which genes are responsible. Researchers at Walailak University, Prince of Songkla University, and Ahvaz Jundishapur University of Medical Sciences set out to map the bacteriocin-associated loci in these strains and compare them across the species.
The team assembled 44 nonredundant L. paracasei genomes after filtering 67 candidates by average nucleotide identity. They then ran pangenome, antiSMASH, and phylogenetic analyses to see which bacteriocin genes are shared across the species and which are restricted to a few lineages.
ProteinIQ in action
Genome mining produces candidate peptide sequences, but it says nothing about what those peptides look like. To get a first view of their shape, the authors picked three representative candidates from distinct genomic contexts: a conserved pair from a small Blp-family locus, a conserved pair from the large multipeptide locus, and ECDLDI_01475, a rare 59-residue accessory peptide found only in WU2302 and two other genomes.
They ran peptide structure prediction with AlphaFold 3 on Google's AlphaFold Server. Predicted models of short peptides can carry strained or unrealistic backbones, especially when confidence is modest, so the team checked each model's stereochemistry before describing it. The methods describe that step directly:
The stereochemical quality of each predicted model was evaluated by generating Ramachandran plots using Protein IQ software v1.0.0.
ProteinIQ's Ramachandran plot tool classified every residue's φ/ψ backbone angles as favored, allowed, or outlier, and flagged glycine and proline residues separately because of their distinct conformational preferences. The resulting plots appear as panels C, F, and I in the paper's Figure 6, next to the ribbon models and electrostatic surfaces the authors calculated in UCSF Chimera. Fuller checks of clashes and rotamers are available in MolProbity.
The authors read these plots together with AlphaFold's pTM confidence scores. Good local geometry combined with a low pTM told them the backbone was plausible residue by residue while the global fold was still uncertain. On that basis they recommend energy minimization and molecular dynamics simulation before drawing conclusions about how these peptides act on bacteria. Per-residue confidence, explained in the pLDDT guide, is a useful companion to the global pTM for short peptides like these.
Key findings
The large bacteriocin-associated locus had a nine-component backbone that was complete in 34 of 44 genomes (77.3%). The small Blp-family locus appeared in 18 of 44 genomes (40.9%). The pangenome contained 7,213 gene clusters, of which 1,848 were core. The accessory candidate group_2723 was identical in sequence and position across the three genomes that carry it, and it showed the strongest phylogenetic clustering of any trait tested (p = 0.0002).
The structure prediction checks gave a mixed but informative picture. The Blp-family peptide formed an extended α-helix with 98.8% of residues in favored regions, one outlier at Tyr19, and a pTM of 0.62. The large-locus peptide folded into a compact helix bundle with 95.7% favored residues, no outliers, and a pTM of 0.34. The rare WU2302 candidate formed an elongated helix with 91.1% favored residues, no outliers, and a pTM of 0.41.
Taken together, the study describes a shared bacteriocin architecture in L. paracasei with a conserved backbone plus lineage-restricted extras, and it narrows the search to a short list of peptides for lab testing against MRSA. The authors note that MIC values were not measured, so the next step is functional validation of these specific candidates.
