Comparative study by adding bootstrapping stage in construction of biological networks

Model selection methods are very popular in high-dimensional settings in recent years due to the availability of massive amounts of data, specifically from genetical, image progressing, and financial sources. Therefore, the selection of the best estimated model becomes crucial. There are a number of model selection approaches in order to choose the optimal one among alternatives. Among them are the Akaike information criterion, Bayesian information criterion, Consistent Akaike information criterion with Fisher information matrix (CAICF), and Information and COMPlexity (ICOMP), which are very successful in lasso regression when constructing biological networks. In this study, we have proposed these criteria by inserting both non-parametric and Bayesian bootstrap approaches to optimize CAICF and ICOMP selection criteria when the sample size is smaller than the number of genes in the system. We evaluate the performance of the bootstrapping strategy with distinct Monte Carlo scenarios. From the majority of results it is shown that the model selection with bootstraps has higher accuracy than the model selection without bootstraps.