Two previously uncharacterised proteins from the plant-associated bacterium Kosakonia cowanii F52 have been assigned tentative identities through a purely computational analysis, and the results now stand as specific, testable hypotheses rather than confirmed functions. One protein, MHQ1860809.1, is predicted to adopt a canonical Rossmann fold with a conserved NAD(P)-binding pocket, and molecular docking of NAD+ into that pocket yielded a Vina score of −7.8 kcal/mol. Gene Ontology predictions for the same protein point to oxidoreductase activity (GO:0016646) and a role in proline metabolic process (GO:0006560). The second protein, MHQ1862734.1, is predicted to fold into an ARM-repeat superhelix, lacks any detectable ligand-binding cavity, and is structurally similar to importin-β and AP-2 adaptor subunits, yet no enzymatic function could be assigned to it. The study, published in Genomics & Informatics, applied established structure-prediction and annotation tools to a genome for which no curated functional database exists, and its authors emphasise that every functional assignment presented is computational and requires experimental validation.
Kosakonia cowanii is a Gram-negative member of the Enterobacteriaceae that has emerged as a multifunctional plant growth-promoting rhizobacterium. It is documented to enhance crop salinity tolerance, suppress fungal pathogens, solubilise phosphate, and produce phytohormones, serving as both a biostimulant and a biofertilizer. Unlike model plant growth-promoting bacteria such as Pseudomonas or Bacillus, K. cowanii lacks a dedicated curated molecular database, and its beneficial traits have been described almost exclusively at the physiological level. Most earlier studies reported increases in biomass or reductions in disease without identifying the proteins or molecular mechanisms responsible, a gap that hinders the rational design of robust bioinoculants to mitigate the impact of soil salinisation on crop yields. The fully sequenced strain F52, whose genome is deposited under GenBank accession CM160069.1, offered the research team an opportunity to probe the molecular machinery behind these traits.
A major obstacle to decoding plant growth-promoting activity is the prevalence of hypothetical proteins, sequences with no experimentally or computationally assigned function. In the F52 genome, which spans 4,708,407 base pairs and encodes 4,261 protein-coding sequences, 5.1 percent of the proteins are annotated as hypothetical proteins, amounting to 218 sequences. These uncharacterised sequences could harbour novel enzymes, signalling adaptors, or effector proteins that directly mediate plant growth promotion and stress tolerance. However, traditional homology-based tools such as BLAST fail for most hypothetical proteins because they lack detectable sequence similarity to experimentally validated entries, leaving them in what the researchers describe as a permanent functional blind spot. The very proteins that might be most unique to K. cowanii F52, and most responsible for its halotolerant traits, remain invisible to conventional annotation pipelines.
The computational pipeline that produced the two candidates relied on sequential filters. Physicochemical properties were evaluated with the ExPASy ProtParam tool, and proteins with an instability index greater than 40 were discarded, reducing the 218 hypothetical proteins to 111 stable ones. Subcellular localisation was then predicted for those 111 stable proteins using four independent tools, CELLO v2.5, PSORTb, SOSUIGramN, and PSLpred, and only proteins for which all four tools returned identical predictions were retained, leaving 11 candidates. Conserved domains were searched in those 11 candidates using the NCBI Conserved Domain Database batch search and InterProScan, which integrates databases including Pfam, PROSITE patterns, SUPERFAMILY, and Gene3D, with a significance threshold of an E-value below 0.01. After removing seven proteins that had no significant domain, four candidates remained: MHQ1859537.1, MHQ1860643.1, MHQ1860809.1, and MHQ1862734.1.
Template-based homology modelling through the SWISS-MODEL server was attempted first, but although experimental templates were identified for all four proteins, the highest sequence identity to an experimentally determined structure ranged from 14.1 to 21.2 percent, far below the 80 percent threshold required for reliable homology modelling. De novo prediction was therefore performed using AlphaFold2 through the ColabFold environment, run without external templates and with AMBER relaxation applied to the top-ranked model. The relaxed models were validated with four standard quality-assessment tools, PROCHECK, ERRAT, Verify3D, and QMEANDisCo, against acceptance thresholds of more than 90 percent most favoured Ramachandran regions, an ERRAT score above 80, more than 50 percent of residues with a 3D-1D score of at least 0.1, and a QMEAN Z-score above −4. MHQ1862734.1 passed all thresholds with PROCHECK at 91.6 percent, ERRAT at 100, Verify3D at 52.17 percent, and a QMEAN Z-score of −0.29. MHQ1860809.1 passed PROCHECK at 92.6 percent and QMEAN with a Z-score of −1.64 but had borderline ERRAT at 75.2 and Verify3D at 30.25 percent, and it was retained because its AlphaFold2 pLDDT confidence is above 90 across the central Rossmann fold core, with low confidence confined to the N- and C-terminal tails. The other two candidates, MHQ1859537.1 and MHQ1860643.1, were discarded.
The physicochemical profiles of the two selected proteins differ in detail. MHQ1860809.1 contains 162 amino acids with a molecular weight of 18,433.21 Da, a theoretical isoelectric point of 5.63, an instability index of 39.89, and a GRAVY hydropathicity value of 0.083, with 19 negatively charged residues and 14 positively charged ones and an aliphatic index of 108.89. MHQ1862734.1 contains 161 amino acids with a molecular weight of 18,055.40 Da, a theoretical isoelectric point of 4.72, an instability index of 38.49, and a GRAVY value of −0.164, with 27 negatively charged residues and 17 positively charged ones and an aliphatic index of 109.25. Both proteins were consistently predicted to be cytoplasmic by all four subcellular localisation tools, and neither sequence carries a recognisable signal peptide or transmembrane helix, which is consistent with an intracellular location. Secondary structure predictions from PSIPRED and SOPMA showed similar distributions for each protein, with SOPMA assigning 71.43 percent α-helices to MHQ1862734.1 and 44.44 percent to MHQ1860809.1.
Conserved domain searches reinforced the structural picture. For MHQ1860809.1, InterProScan identified a NAD(P)-binding Rossmann fold domain, SUPERFAMILY SSF51735 with an E-value of 3.53 × 10⁻⁵ and Gene3D G3DSA:3.40.50.720 with an E-value of 2.4 × 10⁻⁶, while NCBI CD-Search detected a multi-drug resistance superfamily domain, cl16912, with an E-value of 5.85 × 10⁻³. For MHQ1862734.1, CD-Search returned no significant domain at the chosen threshold, but InterProScan revealed an ARM repeat of the Armadillo-type fold, SUPERFAMILY SSF48371 and IPR016024, with an E-value of 5.76 × 10⁻⁷. BLASTp searches against the non-redundant protein database found that the top ten matches for MHQ1860809.1 came from Kosakonia and closely related Enterobacteriaceae, sharing 95 to 100 percent sequence identity and mostly annotated as hypothetical proteins, although several hits from Klebsiella, Serratia, and unclassified Kosakonia were annotated as NAD(P)-binding domain-containing proteins or NADP oxidoreductases. For MHQ1862734.1, the closest homologues were found within Kosakonia cowanii itself, with the top hit sharing 100 percent identity, and all top ten hits were annotated as hypothetical proteins.
Structural similarity searches against the Protein Data Bank using Foldseek identified 761 structurally related proteins for MHQ1860809.1 and 242 for MHQ1862734.1. The highest TM-score for MHQ1860809.1, 0.548, was observed with pyrroline-5-carboxylate reductase from Bacillus cereus, and the second and third hits were also pyrroline-5-carboxylate reductases, from Coxiella burnetii and Medicago truncatula, with TM-scores of 0.506 and 0.505. Sequence identities across the top ten hits ranged from 5 to 9 percent, and the researchers note that these TM-scores and identities indicate fold-level similarity only and do not establish functional conservation. Pyrroline-5-carboxylate reductase catalyses the final step of proline biosynthesis, the NADPH-dependent reduction of Δ1-pyrroline-5-carboxylate to proline, and proline is a well-established osmoprotectant that accumulates in both bacteria and plants under salt stress. For MHQ1862734.1, the highest TM-score, 0.603, was observed with the integrator cleavage module INTS4 from Homo sapiens, with importin beta and AP2 clathrin adaptor mu2 core among the other hits, and no enzymatic domains appeared among the top ten matches. The authors state that functional inference from these cross-kingdom structural matches is not warranted.
Functional annotation through the COFACTOR server predicted oxidoreductase activity acting on the CH-NH group of donors with NAD or NADP as acceptor for MHQ1860809.1 with a confidence score of 0.58, along with proline metabolic process, glutamine family amino acid biosynthetic process, and carboxylic acid biosynthetic process, each at 0.58, and cytoplasmic localisation at 0.73. No confident Enzyme Commission number could be assigned to either protein, with the top five EC predictions for each carrying very low confidence scores, and no significant KEGG Orthology number could be assigned to either protein by the BlastKOALA server. COFACTOR did identify a putative NAD(P)-binding pocket in MHQ1860809.1, with the top-ranked ligand-binding site template being a pyrroline-5-carboxylate reductase crystallised with NADP+, and predicted binding site residues at positions 8 through 12, 54, 67 through 69, 91, and 92. PrankWeb detected a single binding pocket in MHQ1860809.1 comprising 18 residues, but no ligand-binding pocket was detected in MHQ1862734.1, and COFACTOR likewise identified no significant ligand-binding pocket there, observations the authors say are consistent with a non-catalytic, scaffolding role for the ARM-repeat protein.
The docking evidence carries explicit caveats. Molecular docking of NAD+ into the relaxed AlphaFold2 model of MHQ1860809.1 using CB-Dock2 detected five potential binding cavities, with the top-ranked cavity yielding the −7.8 kcal/mol Vina score and a cavity volume of 501 ų, and the interacting residues match the Rossmann motif that anchors the adenine ring. The authors state that the docking result should be interpreted as consistent with, but not proof of, NAD+ binding, that Vina scores are empirical scoring values and not equivalent to binding free energies, and that no negative control, molecular dynamics simulation, or MM/PBSA calculation was performed. Canonical pyrroline-5-carboxylate reductases are approximately 300 amino acids and dimeric, whereas MHQ1860809.1 is only 162 amino acids, although the catalytic Rossmann fold domain of the enzyme is itself approximately 150 to 170 amino acids, suggesting the protein corresponds to the catalytic core rather than the full-length enzyme. No genetic knockout or plant inoculation experiments were performed, so the physiological relevance of both proteins for salt tolerance and plant growth promotion has not been directly tested, and the authors propose targeted knockout mutants, interaction-partner searches for the ARM-repeat protein, and plant inoculation assays under saline conditions as the next steps.
Subject of Research: Computational structural and functional annotation of two hypothetical proteins from the plant-associated bacterium Kosakonia cowanii F52
Article Title: In silico structural and functional characterisation of two hypothetical proteins from the plant‑associated bacterium Kosakonia cowanii F52
Article References: Islam, M. Z., Dewan, M. J. H., Keyhani, A. B., & Fayaz, M. (2026). In silico structural and functional characterisation of two hypothetical proteins from the plant‑associated bacterium Kosakonia cowanii F52. Genomics & Informatics, 24(1), Article 17. https://doi.org/10.1186/s44342-026-00080-9
Image Credits: AI Generated
DOI: 10.1186/s44342-026-00080-9
Keywords: Kosakonia cowanii, hypothetical proteins, AlphaFold2, Rossmann fold, molecular docking, proline biosynthesis, ARM-repeat proteins, plant growth-promoting rhizobacteria, protein structure prediction, functional annotation, salt stress, bioinformatics
News Source: Gregory Coleman. (October 28, 2026). Computer models point to a proline enzyme and a mystery scaffold in a salt-tolerant plant microbe. Scienmag.
