Turning angle analysis reveals hidden anisotropies in the anomalous diffusion of molecules in live cells

What the math gives to ML

The paper provides a rotation-robust diagnostic for detecting anisotropy in short, noisy trajectories: turning-angle statistics retain information about the ratio of intrinsic diffusion coefficients even when the laboratory-frame orientation is random. Its transferable asset is a local geometric statistic that separates temporal anomalous-diffusion correlations from hidden directional structure without estimating a global axis. In neural networks, this can become an anisotropy-aware loss and monitoring signal for trajectory predictors, latent world models, recurrent state-space models, or diffusion samplers. The strongest initial use is to test whether a learned model reproduces turning-angle distributions while remaining invariant to arbitrary global rotations.