Comparative performance of 4 circadian detection algorithms for pooled RNA-sequencing data without replicates

Bulk RNA-sequencing of time course data is a powerful tool for characterizing circadian transcriptional rhythms. However, the cost of sequencing multiple biological replicates across dense time courses often limits feasibility, particularly in exploratory or resource-constrained settings. This study aimed to systematically evaluate the performance of 4 circadian rhythmicity-detection algorithms applied to non-replicated RNA-seq data: MetaCycle, RAIN, DiffCircadian, and a Fourier spectrum–based method with permutation-derived significance. We hypothesized that DiffCircadian would demonstrate increased sensitivity without compromising specificity, given its prevalent use in this field. To test this, we simulated RNA sequencing data. We generated synthetic “pseudo-genes” representing 5 expression patterns including circadian, flat, drifting, random, and non-circadian oscillatory, across 13 evenly spaced timepoints over 48 hours. Observed counts were drawn from a negative binomial distribution, incorporating size factors to model variability in sequencing depth. Amplitudes were varied from 0.8×SD to 1.5×SD to simulate both signal-to-noise differences and overdispersion typical of RNA-seq experiments. Specificity was high (>96%) across all methods, whereas sensitivity increased with amplitude. At very low amplitudes (0.8×SD) MetaCycle detected 13.8% of true-positive rhythms, RAIN 6.8%, DiffCircadian 0.1%, and Fourier 0%. At amplitude 1.5×SD, MetaCycle reached 67.3% sensitivity, RAIN 63.0%, DiffCircadian 43.6%, and Fourier 33.1%. This hierarchy of algorithm performance including accuracy, and precision, was consistent across amplitudes. All algorithms performed better at higher signal-to-noise ratio. These results underscore the limitations of current rhythmicity-detection methods in non-replicate RNA-seq data and highlight the need for more sensitive approaches. All approaches offered highly specific (>97%) conservative detection. However, MetaCycle and RAIN outperformed DiffCircadian and Fourier-permutation approaches. These findings provide practical guidance for performing data analysis in resource constrained settings as well for optimizing circadian experimental design. Funding: NIH/NDDK R25DK13432. Research and Education in Nephrology for Undergraduate Medical Students–Florida (RENUM-FL). NIH/NIDDK 5U24DK128851-02. Innovative Science Accelerator Program (Via Augusta University (Subaward 36350-9)) “Sex differences in the kidney circadian clock mechanism” This abstract was presented at the American Physiology Summit 2026 and is only available in HTML format. There is no downloadable file or PDF version. The Physiology editorial board was not involved in the peer review process.