Sign Gradient Descent Algorithms for Kinetostatic Protein Folding

TL;DR

This paper introduces a sign gradient descent algorithm within the kinetostatic compliance method to efficiently predict protein folding structures, improving convergence speed and robustness over traditional methods.

Contribution

The paper presents a novel SGD-based algorithm for protein folding prediction under KCM, offering faster and more reliable convergence to local energy minima.

Findings

Demonstrates effective folding simulations of protein backbone chains.

Shows improved convergence speed compared to conventional methods.

Validates robustness of the proposed algorithm in folding dynamics.

Abstract

This paper proposes a sign gradient descent (SGD) algorithm for predicting the three-dimensional folded protein molecule structures under the kinetostatic compliance method (KCM). In the KCM framework, which can be used to simulate the range of motion of peptide-based nanorobots/nanomachines, protein molecules are modeled as a large number of rigid nano-linkages that form a kinematic mechanism under motion constraints imposed by chemical bonds while folding under the kinetostatic effect of nonlinear interatomic force fields. In a departure from the conventional successive kinetostatic fold compliance framework, the proposed SGD-based iterative algorithm in this paper results in convergence to the local minima of the free energy of protein molecules corresponding to their final folded conformations in a faster and more robust manner. KCMbased folding dynamics simulations of the backbone chains of protein molecules demonstrate the effectiveness of the proposed algorithm.