Automated database generation and electron density analysis for NOS bond identification

Nitrogen-oxygen-sulfur (NOS) linkages between lysine and cysteine residues represent novel chemical patterns crucial for cellular redox regulation and signalling. Despite their significance, automated identification of these linkages in protein structures remains challenging. Here, we present an algorithm that integrates geometric and electron density screening to detect likely NOS bonds in protein structures. We use a combination of two approaches: a Mid-point approach and a quantum mechanics-informed (QM-point) approach, each applied with varying search radii. We evaluated the algorithm using two datasets: one containing structures which upon manual inspection are confirmed as likely NOS candidates, and a control set where although the N-S distances are within established thresholds, there are is no evidence of NOS formation. The Mid-point method demonstrated strong performance across different search radii, with success rates ranging from 48% to 55% for likely NOS structures. The QM-point approach showed high specificity (99% success rate) for unlikely-NOS structures at a 2.0 Å radius. We propose a two-step screening process that uses the strengths of both methods to optimize NOS bond detection. This approach is an initial step towards automated identification of chemical patterns in protein structures, potentially uncovering previously overlooked linkages and contributing to a deeper understanding of chemical bonds in protein structures.