Identification and characterisation of variable rhythms using CosinorPy

Abstract

AbstractBackground and objectiveAnalysis of rhythmic data has become an important aspect in biological and medical science, as well as other fields of science. Parametric methods based on trigonometric regression reflect several advantages in comparison to their alternatives. Software packages for parametric analysis of rhythmic data are mostly based on a single-component cosinor model, which is not able to describe asymmetric rhythms that might reflect multiple peaks per period. Moreover, a basic cosinor model is unable to describe rhythmicity that is changed with time. Here, we present some important extensions of the recently developed CosinorPy Python package to address these gaps.MethodsThe extended package CosinorPy provides the functionalities to perform a detailed individual or comparative analysis of rhythms reflecting (1) multiple asymmetric peaks per period, (2) forced, damped, or sustained rhythms, and (3) shift of the midline statistics of rhythm over time. In all these cases the package is able to assess the (differential) rhythmicity parameters, evaluate their significance and confidence intervals, and provide a set of publication-ready figures.ResultsWe demonstrate the package in some typical scenarios that incorporate different types of rhythmic dynamics, such as asymmetric, damped, and forced rhythms. We show that the proposed implementation of a generalised cosinor model is capable of reducing the error of estimated rhythmicity parameters in cases that tend to be problematic for alternative models. The implementation of the presented package is available together with scripts to reproduce the reported results athttps://github.com/mmoskon/CosinorPy.ConclusionAccording to our knowledge, CosinorPy currently presents the only implementation that is able to cover all the above scenarios using a parametric model with vast applications in biology and medicine as well as in other scientific domains.