Skip to content

🧬 DiffBiophys

Welcome to DiffBiophys, a high-performance Python library for differentiable biophysical modeling.

Built on JAX, it re-implements core structural biology and spectroscopy observables (SAXS, NMR, CD) as hardware-accelerated, auto-differentiable kernels.

Why Differentiable?

Traditional biophysics libraries provide "forward models" (Structure -> Observable). DiffBiophys provides the gradient, enabling "inverse modeling" (Observable -> Structure).

  • Refine protein loops directly against experimental NMR data.
  • Fit structure ensembles to SAXS scattering curves.
  • Integrate physical constraints into machine learning loss functions.