Optimal dimensionality selection for independent component analysis of transcriptomic data

Independent Component Analysis (ICA) is an unsupervised machine learning algorithm that separates a set of mixed signals into a set of statistically independent source signals. Applied to high-quality gene expression datasets, ICA effectively reveals the source signals of the transcriptome as groups of co-regulated genes and their corresponding activities across diverse growth conditions. Two major variables that affect the output of ICA are the diversity and scope of the underlying data, and the user-defined number of independent components, or dimensionality, to compute. Availability of high-quality transcriptomic datasets has grown exponentially as high-throughput technologies have advanced; however, optimal dimensionality selection remains an open question. Here, we introduce a new method, called OptICA, for effectively finding the optimal dimensionality that consistently maximizes the number of biologically relevant components revealed while minimizing the potential for over-decomposition. We show that OptICA outperforms two previously proposed methods for selecting the number of independent components across four transcriptomic databases of varying sizes. OptICA avoids both over-decomposition and under-decomposition of transcriptomic datasets resulting in the best representation of the organisms underlying transcriptional regulatory network.