Orateur
Description
Predicting the three-dimensional folds of proteins is a celebrated problem at the crossroads of molecular dynamics and artificial intelligence. Our novel approach sets aside amino-acid sequence information and focuses instead on the geometry of protein backbones. An empirical density on SO(3) is determined from experimental protein structures, and a Boltzmann-like construction yields approximations to free energies directly from backbone geometry. This provides a new approach to predicting protein function from structure, with diverse applications across biophysics and biology. Two new physico-geometric quantities are introduced: Fragility, which measures the intrinsic susceptibility of a protein to perturbation, and Lability, which measures its response to a specific perturbation. A remarkable empirical law emerges: regions of high native backbone Fragility tend to exhibit greater Lability under structural perturbations, reminiscent of the fluctuation–dissipation theorem of statistical mechanics.