Information Theory, Inference & Learning Algorithms
Information Theory, Inference & Learning Algorithms
Linear and nonlinear ICA based on mutual information: the MISEP method
Signal Processing - Special issue on independent components analysis and beyond
Adaptive extremal optimization by detrended fluctuation analysis
Journal of Computational Physics
Clustal W and Clustal X version 2.0
Bioinformatics
Bioinformatics
Brief communication: Computation of mutual information from Hidden Markov Models
Computational Biology and Chemistry
Information-theoretic analysis of molecular (co)evolution using graphics processing units
Proceedings of the 3rd international workshop on Emerging computational methods for the life sciences
Hi-index | 0.01 |
Mutual information (MI) is a standard measure in information theory to observe and quantify correlated signals and events in both, empirical data sets and theoretical models. In the field of computational biology the MI turned out to be particularly useful in studies on co-evolutionary signals of sites within biomolecules. A key issue in the applicability of the MI is, however, a correct reference system or null model to understand finite-size effects in the underlying, finite data set. Although some bioinformatics studies exist with rigorous results for theoretical, well-designed random distributions, data from real-world proteins was never used to quantify the effect of finite-size samples. The impact of real-world statistics is, however, most relevant for researchers in all fields concerned with detecting evolutionary signals within biological sequences. We present results on such effects in finite-sized biological data sets and point to future research directions. We are most of all concerned with bacterial, ribosomal proteins as a prototypical example in molecular evolution. We compare to previous published suggestions, give an empirical formula, and propose a protocol to guide future research projects.