The choice of structural template can create apparent links between predicted protein stability and longevity in tetrapods

Abstract

Comparative studies increasingly score amino-acid variation across species with structure-based predictors applied to a single structural template, but whether the template itself can create associations with species traits has received little attention. We examined this while testing whether long-lived species carry proteins predicted to be more stable or, under nearly neutral theory, less stable. In two independent tetrapod radiations, 57 placental mammals (1,755 genes) and 40 reptiles and amphibians (1,910 genes), we reconstructed ancestral proteins, predicted with ThermoMPNN the folding effect of every amino-acid difference between the ancestor and each species on AlphaFold backbones, and related the mean effect to lifespan by phylogenetic regression, following analysis plans fixed in advance. With a human backbone, primates, which are long-lived for their size, appeared least destabilized, producing a weak apparent association with lifespan. In a pre-specified confirmatory test, a platypus backbone reduced the primate offset from -1.96 to -0.62 SD and the association disappeared (P = 0.701). In reptiles, a human backbone, although equidistant in time, favoured slowly evolving turtles and crocodilians and gave a nominal association that two fish backbones did not reproduce. The predictor captured site-level stability signal, yet with outgroup templates neither radiation showed a robust association between predicted stability and longevity, although moderate effects cannot be excluded. The choice of structural template can therefore create apparent cross-species associations. We propose four checks to detect them: ancestral references, outgroup templates, tests for clade and rate effects, and tests of dependence on a few clades.