Differentiable Spin-Center Diffusion Model Enables DEER-Guided Generative Pipeline of Weighted Protein Ensembles

Spin-center diffusion with a tunable spin-label rotameric width enables DEER P(r) to directly guide generation of weighted protein conformational ensembles, recovering bimodal populations.

• Why it matters: DEER distance distributions mix conformational heterogeneity with label rotamer flexibility, and existing label models overestimate the latter, leaving P(r)-to-ensemble translation indirect and mostly qualitative.
• What they did: The authors built a differentiable spin label model (SC diffusion) coupled to a frozen Boltz-2 structure generator, then used experimental P(r) at inference time to generate ensembles and derive condition-dependent free energy differences.
• The result: The pipeline robustly recovers known ensemble populations from synthetic P(r) and models DEER-guided ensembles for PfMATE, LmrP, MhsT, and P-glycoprotein under different biochemical conditions.

AI summary of the abstract · may contain errors · not medical advice · read the paper.