Blind Analysis of Molecular Dynamics

We describe a nonparametric approach for accurate determination of the slowest relaxation eigenvectors molecular dynamics. The is blind as it uses no system specific information. In particular, does not require functional form with many parameters to closely approximate eigenvectors, e.g., linear combinations descriptors or deep neural network, and thus extensive expertise system. suggest rigorous sensitive validation/optimality criterion an eigenvector. only eigenvector time series can be used validate computed by other approaches. power illustrated on long atomistic protein folding trajectories. determined pass validation test at scale 0.2 ns, much shorter than alternative