An improved rhythmicity analysis method using Gaussian Processes detects cell-density dependent circadian...

Area of Science:

• Chronobiology
• Computational Biology
• Genomics

Background:

• Detecting biological rhythms in time-series data is challenging due to low amplitude, high variability, and non-stationarity.
• Existing rhythm detection methods often struggle with these complex datasets and rely on P-values, which can be limiting.

Purpose of the Study:

• To introduce a novel method, Oscillation Detection using Gaussian Processes (ODeGP), for robust detection of biological oscillations.
• To address limitations of current methods by incorporating measurement errors, non-uniform sampling, and non-stationary data.

Main Methods:

• ODeGP combines Gaussian Process regression and Bayesian inference.
• It utilizes a non-stationary kernel and Bayes factors to model both rhythmic and non-rhythmic hypotheses.
• The method is designed for analyzing single or few time-trajectories.

Main Results:

• ODeGP significantly outperforms eight common methods in detecting stationary and non-stationary oscillations using synthetic datasets.
• The method demonstrates higher sensitivity in detecting weak and noisy oscillations in qPCR datasets.
• ODeGP identified rapid Bmal1 gene oscillations in mouse embryonic stem cells upon increased cell density, revealing new biological patterns.

Conclusions:

• ODeGP offers a powerful and sensitive approach for detecting complex biological rhythms.
• The method can uncover previously undetected rhythmic patterns and provide new biological insights.
• ODeGP's ability to handle non-stationary and noisy data makes it valuable for chronobiology research.