prody

ProDy builds Gaussian and anisotropic network models from a structure and computes normal modes from them. For protein-flexibility that makes it the reference line rather than a competitor: a Gaussian network model consumes no training data, fits nothing but one positive scale and offset per protein, and per-protein Pearson correlation is invariant to both — so it cannot leak, and any learned descriptor should be asked to beat it under the same data and split.

It is an enabling dependency of the same shape as biopython-environment and dssp-environment rather than a topological method.

conda-forge, locked green. Bioconda has no prody at all, so no container follows and the fixture grades L3. conda-forge builds 2.6.1 for linux-64 and osx-64 only, and has never built prody for osx-arm64 at any version, so an arm64 macOS reader cannot install this fixture natively whatever the platform set says.

calcModes defaults to twenty modes; squared-fluctuation B-factor prediction needs the whole spectrum, so a caller must ask for it.