Initial implementation of constrained ANM

[anupam-banerjee]

Constrained Anisotropic Network Model (cANM)
A ProDy-compatible extension of the legacy ANM that incorporates higher-order geometric constraints for more physically realistic protein normal-mode analysis.
This package adds several structural terms to the Hessian beyond pairwise distance springs, enabling more accurate modeling of protein flexibility:

1. Pairwise ENM/ANM springs
2. 3-body bond-angle constraints
3. 4-body torsional/dihedral constraints
4. Backbone curvature term

The result is an ANM-like model whose dynamics better reflect local geometry and backbone behavior, while remaining easy to use within the existing ProDy workflow.

It provides:
cANM — a drop-in subclass of ProDy’s ANM
calcCANM(...) — convenience function analogous to calcANM, supporting PDB strings, objects, or raw arrays

This implementation uses dense NumPy arrays (no sparse support).

[AnthonyBogetti]

Hi @anupam-banerjee, can you comment here with a minimal amount of code to run this new addition to ProDy? Just something that if you were to run it, would expect it to work 100% assuming the code is working as intended.

[anupam-banerjee]

@AnthonyBogetti can you please try the following-

from prody import *
import matplotlib.pyplot as plt

pr = parsePDB('input.pdb', subset='ca')

canm = cANM('constrained')
canm.buildHessian(pr, cutoff=15.0, gamma=1.0, k_theta=10, k_phi=1.0, kappa=20)

canm.calcModes(n_modes='all')

resnums = pr.getResnums()
y = calcSqFlucts(canm[3:])
x = resnums
plt.plot(x,y)
plt.xlim(x.min(), x.max())
plt.xlabel('Mode Index')
plt.ylabel('Mean Square Fluctuation')