An information content principle explains regulatory patterns of gene expression across human tissues
Gene expression patterns range from broadly expressed housekeeping genes to highly tissue-specific ones. Notably, many genes exhibit intermediate specificity, characterized by elevated expression in some tissues while low or absent in others. Understanding how regulatory demands scale with tissue specificity offers a valuable opportunity to uncover fundamental principles of genome regulation. By analyzing cis-regulatory element (CRE) counts across human genes with varying tissue specificity, we observed a nonlinear pattern: genes with intermediate specificity harbor the highest CRE count, suggesting distinct regulatory strategies across the expression spectrum. Motivated by this observation, we used the Minimum Description Length (MDL) principle from information theory, together with a maximum parsimony approach from phylogenetics, to quantify regulatory demands across tissues. Our analysis revealed that MDL-based regulatory demand scales consistently with diverse regulatory features, including CRE count, transcription-factor and microRNA targeting, and gene structure. To test whether this scaling changes across the expression spectrum, we partitioned genes by expression breadth. Two patterns emerged: features scaling with MDL in selectively expressed genes tend to act as on/off switches, whereas those in ubiquitous genes serve as fine-tuning knobs. Evolutionary analysis revealed that these regulatory patterns vary with gene age, with alignment between MDL and CRE counts peaking in intermediate-aged genes. Collectively, these results establish MDL combined with maximum parsimony as a powerful framework linking regulatory architecture, expression specificity, and evolutionary age, offering novel insights into the organizational principles underlying genome regulation.