Moderate prevalence and evolutionary causes of paralogous compensation in humans

Paralogous compensation drives genetic robustness; however, its in vivo prevalence in humans and the evolutionary forces shaping it remain unresolved. We CRISPR-screened 678 human paralog pairs to identify pairs with negative genetic interactions (nGIs) or compensatory paralogs, and integrated our results with four studies to assemble two complementary resources: the 2168-pair Union and 4859-pair Ito datasets. We extrapolated to the organismal level by classifying paralogs as core (broadly compensating, 99 Union/120 Ito pairs), conditional (111/566) or non-interacting (1958/4173), and testing for enriched compensatory signals in core pairs relative to the latter two groups. Core nGI pairs consistently showed prenatal co-expression, depletion of disease genes and enrichment for upregulation after paralog loss in cancer, though the signals were often moderate. Evolutionary analyses found similar fractions (∼2%) of core compensators among both ancient and young paralogs. Ancient pairs retain overlaps likely under pleiotropic and dosage constraints. Incomplete small-scale duplication mechanisms limit compensation among young pairs, except X-derived retrogenes, which are overrepresented among core nGIs (∼14%). This is consistent with selection to counteract male germline X inactivation and subsequent out-of-testis expansion. Overall, paralogous compensation is moderately prevalent in vivo in humans, shaped by multiple evolutionary forces acting on ancient or young duplicates.