A new open-access study in BMC Genomics has now revisited the question at unprecedented scale, screening the annotations of 23,435 NCBI reference genomes, comprising 22,606 bacterial and 829 archaeal assemblies. The work, conducted by independent researcher Kiarash Sadeghian Esfahani, goes beyond a simple gene-counting exercise. Its central contribution is an architecture-aware audit that asks not just whether a genome carries a label mentioning NucS, but whether the encoded protein actually contains the complete two-domain structure needed to function as a genuine NucS enzyme.

The technical logic of the study hinges on protein domain architecture. A complete NucS protein must carry both an N-terminal NucS_N domain, captured by the Pfam profile PF21003, and a C-terminal NucS_C domain, captured by PF01939, on the same polypeptide. A canonical MutS protein must match five defined MutS profile hidden Markov models on a single protein, and must not carry a same-protein PF01713 domain, which would suggest a small MutS-related protein rather than the canonical mismatch repair form. MutL, the partner protein, requires the PF01119 and PF08676 profiles together. By applying these profile criteria with Pfam profile HMMs, the study converts fuzzy annotation wording into precise structural verdicts.

The initial annotation screen produced a tantalizing result: 1,145 genomes appeared to harbor both NucS and canonical MutS-MutL, a pattern that, if real, would challenge the 2017 mutual-exclusivity finding. But when the stricter multidomain criteria were applied, only 470 of those genomes passed all three tests. Of these, 421 were Halobacteria, 48 were Deinococcota, and a single assembly, from Lysobacter korlensis, was flagged as highly contaminated and therefore unreliable. The coexistence of both repair systems, it turns out, remains concentrated in exactly the lineages identified nearly a decade ago, with no new coexistence lineage emerging from the expanded dataset.

The most revealing part of the analysis concerns what loose annotation language actually means. Across 6,270 unique NucS candidate proteins, 5,189 carried both the N-terminal and C-terminal profiles, while 111 carried only NucS_N, 901 carried only NucS_C, and 69 matched neither profile. In other words, roughly one in five candidate proteins fell short of the complete two-domain architecture, even though they had been picked up by annotation-based screening.

The discrepancy becomes even starker in the tier of genomes annotated with the generic phrase NucS domain-containing. That tier contained 682 genomes, of which 680 were testable, yielding 717 unique candidate proteins. Among these, 685 were NucS_C-only, eight were NucS_N-only, five carried both profiles, and 19 carried neither. At the genome level, 663 of the 680 testable generic-tier genomes, or 97.50 percent, contained at least one NucS-associated domain, but only five contained a same-protein two-domain NucS candidate. A generic annotation is therefore usually valid at the domain level, yet it is a poor proxy for a complete, potentially functional NucS protein.

By contrast, explicit nucS gene annotations proved far more trustworthy. Genomes with explicit nucS annotations showed 99.63 percent concordance with the two-domain profile requirement, meaning that when annotators commit to naming the gene, the underlying protein almost always carries the full architecture. Canonical MutS and MutL annotation-positive genomes met the strong profile criteria in 96.78 percent and 97.14 percent of cases, respectively. The lesson for comparative genomics is that annotation specificity matters: broad domain-containing labels and precise gene names carry very different evidentiary weight.

To quantify how rarely the two repair systems truly coexist, the study computed the Jaccard overlap between genomes harboring two-domain NucS candidates and genomes meeting the strong MutS-MutL criteria. The result was just 2.18 percent, a figure that underscores how mutually exclusive the two pathways remain across the contemporary reference collection. The Halobacteria, with 421 validated coexistence genomes, and Deinococcus-Thermus, with 48, stand as the enduring exceptions to a rule that now holds across more than twenty-three thousand assemblies.

The author is careful to frame the scope of these conclusions. Because genomes lacking relevant annotations were not exhaustively searched with profile HMMs, the reported counts represent validated positives from an annotation-first workflow rather than exhaustive prevalence estimates or gene-loss reconstructions. Some true NucS or MutS-MutL genes may lurk in poorly annotated genomes that the screen never reached. This distinction matters for anyone hoping to use the numbers as a census of mismatch repair diversity, but it does not undermine the central finding about lineage concentration.

For the field of prokaryotic genomics, the study arrives at a moment when automated annotation pipelines and large language models are increasingly used to infer gene function at scale. The demonstration that a generic domain-containing label can be 97 percent likely to indicate some NucS-associated domain yet only rarely indicate a complete two-domain protein is a cautionary tale for downstream analyses. Whether the goal is reconstructing the evolution of DNA repair pathways, predicting genome stability phenotypes, or curating protein family databases, the message is clear: verify the full multidomain architecture before trusting the label. As reference genome collections continue to grow, architecture-aware validation of this kind may become a standard step in turning annotation databases into reliable biological knowledge.

Subject of Research: Comparative genomics of NucS and MutS-MutL DNA mismatch repair systems across prokaryotic reference genomes

Article Title: Annotation-aware multidomain validation of NucS and MutS-MutL across 23,435 prokaryotic reference genomes

Article References: Annotation-aware multidomain validation of NucS and MutS-MutL across 23,435 prokaryotic reference genomes. (n.d.). https://doi.org/10.1186/s12864-026-13470-8

Image Credits: AI Generated

DOI: 10.1186/s12864-026-13470-8

Keywords: DNA mismatch repair, NucS, EndoMS, MutS, MutL, comparative genomics, Pfam, profile HMM, RefSeq, protein domain architecture, archaea, bacteria

The problem the Canadian team set out to solve is deceptively simple to state and notoriously hard to crack. When a protein’s structure has been solved experimentally, domain boundaries can be traced by eye or by algorithm through the folded chain. But structural determination remains slow, expensive, and unevenly distributed across the tree of life. For everything else, scientists lean on conserved domain annotations from resources such as Pfam, InterPro, and the NCBI Conserved Domain Database, or they resort to iterative truncation screening, building a panel of constructs and testing which ones express, fold, and behave well. Each of these strategies has blind spots. Annotation databases capture only domains that have already been characterized in other proteins, leaving novel or fast-evolving domains invisible. Truncation screening is empirical, laborious, and often burns months of laboratory time on constructs that fail. Seq2DomML was designed to complement these approaches by learning, from data, what a domain boundary looks like in the raw language of sequence.

The foundation of the framework is the Evolutionary Classification of Domains, or ECOD, a curated database that organizes protein domains by their evolutionary and structural relationships. The researchers converted ECOD’s structural domain annotations into binary residue-level labels: each amino acid in a bacterial protein sequence was marked as belonging to an annotated domain, or alternatively as part of a domain boundary, an inter-domain linker, or a terminal flanking region at the ends of the chain. This transformation turns a structural biology question into a supervised sequence labeling problem, the same class of task that powers speech recognition and named-entity extraction in language technology. Every residue becomes a token to be classified, and the model must learn the contextual cues that signal whether a given position sits inside a folded unit or in the flexible connective tissue between them.

Feeding that problem required an unusually rich encoding of each residue. The team initially described every amino acid with 2,389 features spanning amino acid identity, positional information along the chain, physicochemical properties, and interaction-related descriptors. That breadth is a deliberate hedge: no single property cleanly distinguishes a boundary residue from an interior one. Hydrophobicity, conservation patterns, local sequence composition, and position relative to the termini all carry partial signals, and a boundary might be legible only in their combination. But a feature space of nearly 2,400 dimensions also brings the classic curse of dimensionality, inviting overfitting and burying informative signals in noise. The researchers therefore invested heavily in feature selection, and this is where Seq2DomML acquires its most distinctive technical character.

Feature reduction proceeded in two stages. First, correlation-based pruning removed redundant descriptors that carried overlapping information. Then came the evolutionary step: a multi-objective genetic algorithm that treated the feature set itself as a genome to be optimized. In the initialization stage, the algorithm randomly assigned numbers of active features across a population and generated 100 unique binary feature vectors, each one a candidate recipe of which descriptors the model should see. Pairs of parents were then matched at random and combined through two-point crossover to produce two offspring per couple, with bit-flip mutation applied to the offspring to preserve diversity across the population. Each generation, every parent and offspring was trained and evaluated on three objectives simultaneously: the macro F1-score, the prediction rate for the minority boundary class, and the recall for that class. A non-dominated sorting algorithm, borrowed from multi-objective optimization theory, ranked individuals by their best achievable trade-offs across the three objectives, and the top individuals, equal in number to the original population, survived into the next generation. Crossover, mutation, and selection repeated for 50 generations, gradually sculpting a compact feature set that balanced overall accuracy against sensitivity to the rare and difficult boundary residues.

The classifier at the heart of the framework is a bidirectional long short-term memory network, a recurrent architecture well suited to sequential data. Bidirectionality matters here for a reason rooted in protein biology: the signal that a residue sits at a domain boundary depends on context from both sides of the chain, the tail end of one folded unit and the beginning of the next. An LSTM reading the sequence only left to right would encounter boundary evidence after the fact; a bidirectional model integrates information flowing in both directions, allowing the hidden state at each position to reflect the entire local sequence environment. The LSTM’s gating mechanisms, which learn to retain or forget information over long stretches of input, help the network bridge the dozens or hundreds of residues that may separate a boundary from the nearest obvious sequence landmark.

Evaluation was designed to guard against the most common pitfall in protein machine learning: leakage of evolutionary memory. The researchers removed from the held-out test set any sequences sharing at least 35 percent global sequence identity with the training set, ensuring the model could not simply recognize close relatives of proteins it had already seen. Against this stringent benchmark, Seq2DomML was compared with established annotation tools including InterPro, Pfam, and NCBI CDD, and it outperformed them on the classification metrics. It achieved a macro F1-score of 0.734, a figure that weights performance across all residue classes equally and therefore reflects genuine skill on the underrepresented boundary and linker categories, alongside a within-domain F1-score of 0.924, showing near-expert reliability at recognizing residues inside annotated domains. Its domain count mean absolute error of 0.952 was the second-lowest among the evaluated methods, trailing only Pfam.

Perhaps the most telling statistic is the mean domain Intersection over Union, a metric borrowed from computer vision that measures the spatial overlap between predicted and reference regions. Seq2DomML posted the highest mean IoU of the field, 0.846, indicating the strongest spatial agreement with the ECOD-derived domain regions among all methods tested. In practical terms, this means the model does not merely tally the right number of domains; it draws their edges in roughly the right places. For an experimentalist deciding where to cut a protein to isolate a single domain for structural studies, edge placement is everything. A prediction that captures the correct domain count but misplaces a boundary by thirty residues can yield a construct that aggregates, misfolds, or refuses to crystallize. The IoU result suggests Seq2DomML’s boundaries are tight enough to be actionable.

The authors are careful to position the tool as a complement rather than a replacement for the annotation infrastructure that has served biology for decades. Conserved domain databases rest on curated evolutionary evidence and remain authoritative where homologs exist. Seq2DomML’s value emerges precisely where that evidence runs out: in bacterial proteins with no experimentally determined structure, no recognizable domain homologs, and no reliable annotation to guide construct design. In those cases, the model’s residue-level partitions can help prioritize candidate splice regions, flagging the most promising places to truncate a protein before a single clone is made. Because it requires only sequence as input, it scales to the enormous and rapidly growing output of bacterial genome sequencing projects, where the majority of encoded proteins still await structural or functional characterization.

The broader significance of the work lies in its demonstration that structurally defined partitions are recoverable from sequence-derived features alone. The multi-objective evolutionary feature selection adds a methodological lesson for the field: when the class of interest is rare and the feature space is vast, optimizing a single accuracy metric can quietly sacrifice exactly the predictions that matter most. By forcing the feature set to balance macro F1, boundary prediction rate, and boundary recall simultaneously, the Wilfrid Laurier team built a model that stays honest about its hardest task. As structural biology pipelines increasingly depend on computational triage to decide which proteins and which fragments are worth experimental effort, frameworks like Seq2DomML point toward a future in which the chapter breaks of every protein, however novel, can be read straight from its sequence.

Subject of Research: Machine learning prediction of protein domain boundaries from bacterial amino acid sequences

Article Title: Seq2DomML: Protein domain boundary prediction from bacterial protein sequences using a machine learning framework

Article References: Monks, A., Asilian Bidgoli, A., & Suits, M. D. L. (2026). Seq2DomML: Protein domain boundary prediction from bacterial protein sequences using a machine learning framework. BMC Bioinformatics. https://doi.org/10.1186/s12859-026-06666-y

Image Credits: AI Generated

DOI: 10.1186/s12859-026-06666-y

Keywords: protein domains, domain boundary prediction, machine learning, LSTM, feature selection, ECOD, bioinformatics, bacterial proteins, construct design, sequence annotation, Pfam, InterPro