When modelling biological processes, we often find ourselves faced with complex patterns, that cannot be adequately described by existing methods of modelling and data prediction. That applies well when data contains rhythmical elements, as chosen method must be able to process amplitude, period and acrophase variations coupled with smaller data set, which is often biased by human error. With this in mind, it is important to know which methods currently in use can be best suited for such problems. In this work we will focus on few selected methods and their performance in five different synthetically generated data series. With the help of obtained results we will be able to determine which method is better suited for which conditions and performs best in most series, as well as which sampling conditions produce most and least suited data for further computational analyses. We demonstrate the application of selected methods on the analysis of transcriptomic data obtained from GEO database.
|
