The authors reported a new probabilistic algorithm, scRhythmBayes, aimed at resolving cell-level circadian organization directly from single-cell transcriptomes. The model couples a seed-gene count likelihood with a stratum-aware circular prior to infer unobserved cellular circadian phase when sample-level zeitgeber time provides only partial temporal information. The study team observed that existing tools had addressed only selected aspects of single-cell circadian analysis; scRhythmBayes is described as modeling how cell-level phase evidence informs gene-level rhythmic parameters and rhythm-state remodeling. The method assigns posterior probabilities to interpretable rhythm-state categories, including rhythm gain, rhythm loss, amplitude remodeling, and phase shifts. The abstract indicates the algorithm integrates phase support into downstream gene-level rhythmic modeling rather than treating phase estimation and rhythmic testing as separate steps.
Investigators benchmarked scRhythmBayes in heterogeneous simulations and found higher phase-recovery accuracy than existing phase-inference methods, according to the abstract. They reported balanced performance across three tasks: rhythmic gene detection, parameter recovery, and rhythm-state classification. The authors also conducted stress tests to define practical dataset requirements; these tests identified dependencies on seed-gene support, zeitgeber anchoring, temporal sampling density, and cell number. The preprint text does not report numeric thresholds for those requirements in the abstract; detailed parameters and quantitative performance metrics are available only in the full manuscript and supplementary materials referenced by the source.
When applied to mouse liver single-cell data, scRhythmBayes revealed distinct circadian phase occupancy and rhythm-state organization between periportal and pericentral hepatocytes. The study team interpreted these findings as support for a temporal dimension of liver zonation: cells located in different microenvironments showed different phase distributions and remodeling behaviors. The algorithm also identified cell-type-specific rhythmic markers that were largely distinct from conventional static markers, indicating that temporal programs can contribute to cellular identity in ways that average-expression markers do not capture. The abstract does not enumerate the specific marker genes or provide effect sizes; those details are contained in the full preprint and supplementary files noted in the source.
Investigators applied scRhythmBayes to human heart disease data and reported patterns of temporal disorganization that were cell-type selective rather than uniformly distributed across the tissue. Data indicated phase-coordination and phase-distribution analyses can distinguish cell-type-specific temporal remodeling from a global circadian failure. The abstract states the method identified temporal disorganization in particular cell types but does not specify which cell types or the extent of disruption in the summary. Users should consult the main manuscript and supplementary materials for gene-level lists, cell-type annotations, and quantitative assessments.
The authors provided a GitHub link in the article footnotes (https://github.com/Liwei-DynamicsLab/scRhythmBayes) for data and code access. Supplementary material and external data/code links are also referenced in the preprint record. The article is a preprint and has not been certified by peer review; the authors declared no competing interests. The abstract indicates stress tests established practical requirements for successful application—seed-gene support, zeitgeber anchoring, sampling density, and cell counts—but the abstract itself does not list explicit numeric thresholds or parameter settings. Readers and implementers must consult the full manuscript and supplementary files for exact methods, hyperparameters, quantitative benchmarks, and specific gene or cell-type findings.
Limitations and unanswered questions documented in the source: because the report is a preprint it has not undergone peer review; the abstract does not provide detailed numeric performance metrics or the specific seed genes and parameter thresholds used in the stress tests. The source points readers to supplementary material and a GitHub repository for reproducibility resources; those resources should be consulted to evaluate applicability to other tissues, to confirm computational requirements, and to review exact implementation details before deployment in new datasets.