An Entropy-based Statistical Workflow Provides Noise-Minimizing Biological Annotation for

Abstract

This study aims to expand the efficiency of the interpretation concerning the aging process, by exploring a broad gene set, derived from the analysis of an integrative transcriptomic microarray dataset. The dataset comprises human skeletal muscle samples, obtained from healthy males and females, that were used to derive a gene signature of a high informative content, with respect to its functional association with the aging phenotype. Towards this end, a multilayered computational workflow integrating advanced statistical methodologies for the derivation of reliable confidence measures, distribution-based entropy calculations to examine the informational content of the dataset, enrichment analysis, graph-theoretic methods and intuitive visualization was applied. Specifically, statistical testing revealed differentially expressed genes, while an uncertainty calculation algorithm, exploiting Gene Ontology (GO) terms annotations, extended the list of significant genes from 254 to 2791, namely p-value threshold was increased from 0.0005 to 0.103, while keeping simultaneously noise measurements legitimately low. This rich gene set associated functionally the macroscopic phenotype of muscular aging with highly informative, stably correlated with each other, molecular annotations in the GO database. Finally, a set of 57 reliable genes was identified that comprise a gender-independent aging signature, after incorporating crucial information about genes pivotal regulatory role as inferred by the GO tree. The biological interpretation was highly assisted by the illustration of the functional mappings between genes, cellular location and biological processes through circle packing graphs.