A mixed-effects cosinor modelling framework for circadian gene expression

Area of Science:

• Chronobiology
• Genomics
• Biostatistics

Background:

• Circadian rhythms influence molecular and physiological processes, including gene expression.
• Individual variations in circadian clocks (phase offsets) pose challenges for population-level analysis.
• Standard cosinor models can produce biased amplitude estimates due to uncorrected phase offsets.

Purpose of the Study:

• To develop a novel method for estimating population-level cosinor model parameters for gene expression.
• To mitigate attenuation bias in amplitude estimates caused by individual circadian phase offsets.
• To provide a cost-effective alternative to laboratory testing for determining individual phase offsets.

Main Methods:

• Estimating population-level cosinor parameters per gene without individual offsets.
• Estimating individual-level cosinor parameters per person and gene.
• Computing data-driven phase offsets for each individual.
• Re-estimating population-level cosinor parameters incorporating data-driven offsets.

Main Results:

• The proposed method effectively mitigates attenuation bias in population-level amplitude estimates.
• Hypothesis testing for gene expression oscillations shows improved accuracy.
• Simulation studies confirm the method's effectiveness.
• Real-world data application closely matches results from laboratory-based offset determination.

Conclusions:

• This novel method accurately estimates population-level circadian gene expression parameters without expensive lab tests.
• It improves the detection of health-associated gene expression rhythms by correcting for individual phase differences.
• The approach offers a robust and scalable solution for chronobiological research.