Statistical Embedding in Complex Biosystems

Complex high-dimensional systems represent an important area of interdisciplinary research in systems biology. Gene expression values obtained by microarray data represent a good example, owing to their various features that depend on biological network dynamics. This work emphasizes the role of blind source separation for dealing with dimensionality reduction and feature selection, and their useful combination with fuzzy rules, embedding principles and entropic measures. In particular, entropy and embedding are useful tools for controlling the robustness and stability of the decomposition of a system with larger than intrinsic dimensionality. As a result, the convergence to a small intrinsic dimensionality occurs by the means of least dependent components, seen as a minimal number of salient features.