A potassium ion channel simulated with a universal neural network potential

TL;DR

This paper demonstrates that a universal neural network potential can effectively simulate the potassium ion channel's selectivity filter, revealing new molecular insights and transport mechanisms that are difficult to capture with traditional methods.

Contribution

The study applies a universal neural network potential to simulate the potassium ion channel's selectivity filter, uncovering novel hydrogen bonds and transport mechanisms not previously observed.

Findings

Identification of a hydrogen bond stabilizing water in the SF

Observation of carbonyl backbone flipping at new sites

Demonstration of neural network potential's ability to model complex biological systems

Abstract

Potassium ion channels are critical components of biology. They conduct potassium ions across the cell membrane with remarkable speed and selectivity. Understanding how they do this is crucially important for applications in neuroscience, medicine, and materials science. However, many fundamental questions about the mechanism they use remain unresolved, partly because it is extremely difficult to computationally model due to the scale and complexity of the necessary simulations. Here, the selectivity filter (SF) of the KcsA potassium ion channel is simulated using Orb-D3, a recently released universal neural network potential. A previously unreported hydrogen bond between water in the SF and the T75 hydroxyl side group at the entrance to the SF is observed. This hydrogen bond appears to stabilize water in the SF, enabling a soft knock-on transport mechanism where water is co-transported through the SF with a reasonable conductivity (80 $\pm$ 20 pS). Carbonyl backbone flipping is also observed at new sites in the SF. This work demonstrates the potential of universal neural network potentials to provide insights into previously intractable questions about complex systems far outside their training data distribution.