Proteins that reversibly adopt multiple stable tertiary folds challenge the traditional sequence–structure paradigm and are difficult to detect experimentally. The authors introduce Morpheus-3D, a sequence-based framework designed to quantify residue-level tertiary structural diversity and thereby identify proteins with fold-switching potential while localizing the regions responsible for conformational transitions. The method focuses on variation in tertiary interaction environments rather than on secondary-structure predictions alone.
Morpheus-3D computes per-residue entropy profiles derived from the Foldseek 3Di structural alphabet. These entropy profiles measure the diversity of local tertiary interaction environments sampled in structural matches, providing a residue-resolved metric of conformational plasticity. By capturing variation at the level of tertiary interactions rather than secondary structure alone, the approach aims to reveal switching behavior that secondary-structure-based methods can miss.
A central feature of Morpheus-3D is its ability to simultaneously detect proteins likely to switch folds and to localize the sequence stretches that mediate the switch. The entropy profiles highlight regions with elevated structural diversity; these regions correspond to residues inferred to participate in alternative tertiary interaction networks and thus to conformational transitions. The framework therefore supplies both a global fold-switching score for a sequence and a residue-level map that can be used to target experimental follow-up or computational modeling.
According to the authors, Morpheus-3D outperforms existing fold-switching predictors. It accurately recovers experimentally characterized switching regions from known examples and demonstrates improved sensitivity to conformational plasticity that is not captured by methods relying primarily on secondary-structure variation. Details of benchmark datasets, performance metrics, and statistical comparisons are reported in the full preprint; specific numeric performance values or thresholds were not reported in the abstract.
The framework generalizes to recently discovered natural and engineered fold-switching proteins that were absent from the training set. This suggests that entropy profiles based on tertiary interaction variation encode features of conformational plasticity that transfer across sequence space. The authors emphasize that Morpheus-3D can detect forms of conformational diversity inaccessible to secondary-structure-based approaches, extending the reachable discovery space for metamorphic proteins.
Morpheus-3D was applied to 57 representative proteomes to survey the prevalence of fold-switching potential across diverse lineages. The analysis indicates that fold-switching potential is widespread but shows enrichment in lineages associated with regulatory function, pathogenicity, and environmental adaptation. These enrichment patterns suggest biological contexts in which structural plasticity may be selectively maintained or exploited.
To probe evolutionary origins and trajectories, the authors integrated Morpheus-3D predictions with ancestral sequence reconstruction. This combination uncovered evolutionary paths by which conformational plasticity can emerge, providing insight into how fold-switching potential may arise or be lost over evolutionary time. The abstract notes these evolutionary inferences but does not provide detailed ancestral reconstructions in the summary; readers should consult the full preprint for methods and specific examples.
To facilitate use and validation, Morpheus-3D has been implemented as an interactive web platform (https://morpheus.slicearrow.com/). The platform displays per-residue entropy profiles alongside sequence and three-dimensional structure interactively, enabling predicted fold-switching regions to be mapped onto structures and exported for downstream analysis. The authors also provide code and resources: a GitHub repository and a dataset hosted on Hugging Face. These resources are intended to make predictions accessible and reproducible for the community.
This work is reported as a bioRxiv preprint and has not been certified by peer review. The authors declared no competing interests. Funding sources listed include the Wellcome Trust/DBT India Alliance and ANRF India. Links to the GitHub repository, the Morpheus-3D web server, and the dataset on Hugging Face are provided in the preprint.
Morpheus-3D offers a sequence-only, scalable approach to discovering metamorphic proteins and mapping conformationally plastic regions. Because the report is a preprint, independent validation, peer review, and community benchmarking will be important next steps. Users interested in applying Morpheus-3D should consult the full text for implementation details, benchmark datasets, and guidance on interpreting entropy profiles and fold-switching scores.