Conserved fragments redraw how bacteriophage proteins are built
Bacteriophage proteins rarely behave like tidy textbook domains, and this preprint argues that the evolutionary units inside them are more irregular, more nested, and more numerous than standard annotation systems have captured. By building Evolutionarily Conserved Fragments across thousands of phage genomes, the investigators created a sequence-derived framework that reaches beyond classical domain boundaries and tracks conservation at multiple scales. The practical result is a broader map of phage protein organization, one that better reflects how viral proteins are assembled, exchanged, and repurposed during evolution.
The study centered on a very large comparative search. Using HMM-HMM comparisons across proteins from 5,396 RefSeq phage genomes, the team defined 13,035 Evolutionarily Conserved Fragments, grouped into 3,167 families, without forcing the data into preselected domain shapes. That methodological choice matters because bacteriophage proteins often contain mosaics of inherited parts, where a conserved block may correspond to only part of a domain, an entire domain-like unit, or a region that has no counterpart in existing domain libraries. In other words, the analysis was designed to let the proteins define their own modular boundaries rather than inherit them from database conventions.
A modular grammar that sits inside and beyond Pfam and ECOD
The fragment set overlapped substantially with both Pfam and ECOD, yet the overlap was far from complete. ECFs formed a heterogeneous spectrum that included subdomains, domain-equivalent units, composite regions, and fragments with no detectable match in either reference resource. Some relationships were nested, meaning conserved pieces appeared inside larger conserved pieces rather than lining up neatly end to end. In phage biology, where recombination and modular exchange are central to protein evolution, nesting is not a curiosity; it is a structural signature of proteins that are repeatedly cut, recombined, and recomposed across lineages.
A short sentence captures the central point: classical domains do not exhaust the conserved architecture of phage proteins.
Because the fragments were derived from conservation patterns rather than fixed structural assumptions, the resulting map can separate what is truly conserved from what is merely annotated by convention. That distinction is especially useful for viral proteins, where structural prediction alone may miss short but functionally crucial segments, linker-like regions, or motifs embedded inside larger folds. The study therefore adds a layer of resolution between whole-protein annotation and full domain annotation, a scale that is often invisible in routine pipelines.
Broader protein coverage and stronger gains in infection-related functions
One of the clearest findings is coverage. ECFs annotated 50% of phage representative proteins, at least doubling the coverage achieved by either database alone. The gains were largest in structural proteins and infection-related functions, which fits the biology of phage entry and assembly. Proteins involved in host recognition, capsid organization, tail construction, and infection mechanics often accumulate modular innovations because those features are under strong evolutionary pressure to change without breaking core function. Conserved fragments can capture those constrained regions even when the broader protein family looks too divergent for standard domain calls.
The same pattern held in independent metagenomic data, strengthening the case that the fragment framework is not merely an artifact of the RefSeq set. Function-dependent patterns of diversity and spatial segregation also emerged within ECF families, suggesting that some conserved segments are concentrated in particular parts of a protein while others are more broadly distributed. In a viral genome, this kind of positional bias can reflect how different segments are inherited, how they tolerate substitution, or how they contribute to interactions with host receptors and other viral components.
The biology is straightforward in principle and messy in practice. A receptor-binding protein may keep one structural scaffold while reshaping the surfaces that contact bacterial receptors, and a fragment-level system is better suited than a whole-domain label to capture both conservation and diversification inside that protein.
Receptor-binding proteins expose the limits of domain-only annotation
Receptor-binding proteins received special attention because they often combine recognizable folds with poorly annotated or ambiguous segments. In these proteins, ECFs subdivided established domains, identified previously unannotated structural elements, and recovered homologous regions even where structural similarity was uncertain. This is an important proof of concept. If the conserved fragment approach can dissect receptor-binding proteins, it can also help interpret the parts of phage proteins most directly tied to host range, adsorption, and early infection steps.
A second short sentence follows naturally from the data: fragment-level annotation recovered information that standard domain catalogs left behind.
The study also carried an analytical dividend beyond the named proteins. ECFs provided functional associations for 23,925 protein families that lacked detectable Pfam or ECOD domains. For a field in which many viral proteins remain annotated as hypothetical or incomplete, that level of functional extension has practical value. It gives researchers a way to infer relationships among otherwise opaque phage proteins and to organize follow-up work around conserved sequence units that may correspond to reusable biological modules.
A sequence-first framework for phage evolution
The main conceptual advance is not simply more annotations. It is a shift in scale. Phage proteins appear to be built from evolutionary building blocks that do not always match classical domains, and those blocks can operate at several organizational levels at once. Some are domain-sized, some are smaller than domains, and some span composite regions that make more sense in the context of phage evolution than in the context of general protein taxonomy. That is a useful correction for viral protein biology, where modular exchange is a dominant evolutionary force and where conservation may be distributed across fragments rather than concentrated in canonical units.
For computational biology, the resource offers a complementary annotation system that can be layered onto existing databases rather than replacing them. For virology, it offers a way to think more precisely about how complex bacteriophage proteins are assembled and diversified. For functional genomics, it offers a larger target space for experimental validation, especially in proteins that remain outside the reach of current domain libraries.
The work is still a preprint and has not been peer reviewed, so the resource will need external scrutiny before it becomes a standard annotation layer. Even so, the evidence presented here makes a strong case that bacteriophage proteins are organized by conserved sequence fragments at multiple scales, and that those fragments capture functional and structural information that domain-centric approaches miss. The remaining question is not whether phage proteins are modular; it is how many of the modules have been invisible to the annotation systems used so far.