Piecewise regression analysis of biological data with parallel processing

Although many methods exist for analyzing biological data, some have had limited practical utility because of poor speed performance with larger data sets (e.g., protein folding models, reconstructing phylogenies). Piecewise Regression Analysis (PRA) is a downloadable Java application with a user-friendly graphical user interface that takes advantage of multi-core architecture in personal computers to determine whether a data set is modeled more accurately by linear regression or a two-segment piecewise regression. Regression values for two-segment piecewise regression are obtained by searching parameter space with a differential evolution algorithm. Performance is enhanced with parallelization using the ForkJoin library. Rigorous testing of simulated data with varying degrees of variance and number of observations has ensured accurate output. Improvements in speed performance have been confirmed by testing the application on different computers, and with different numbers of processors used for parallel processing. A demonstration of this program is presented with the analysis of zebrafish developmental data.