Integrated genome-scale approach to map elemental proteome interactions

Metal ions such as iron, zinc, copper, and manganese are essential to all life. In bacteria, they enable growth, regulate metabolism, and control the ability to infect a host and cause disease. But the roles and binding sites of these elements within bacteria remain poorly understood as they are readily lost during laboratory techniques such as protein purification.

Current processes used in metalloproteomic investigations are labour- and time-intensive. This project develops an advanced computational platform that can deliver high-confidence predictions of metal-binding protein sites across entire bacterial genomes by integrating advanced artificial intelligence with experimentally validated metalloproteomic data

The platform is a structural prediction tool that can provide insight into metal-binding capabilities and specificity of bacterial pathogens. It combines two existing artificial intelligence models using AlphaFold3-predicted proteins and validated metalloprotein structures from the Protein Data Bank. The first model is a graph-based neural network that locates metal-binding residues. The second model is a 3D pocket detector that evaluates the local protein environment to determine which metal ion fits each site.

This deep learning model will then be refined by comparing predictions against existing databases to deliver calibrated metal-specific binding probabilities. The research team will establish rigorous standards and gold-standard benchmarks for each metal, then validate the pipeline using the genomes of Klebsiella pneumoniae and Streptococcus pneumoniae. Laboratory experiments will validate key predictions by measuring metal content and testing protein activity.

This platform will address major knowledge gaps in metallobiology by:

• improving the capacity to identify and distinguish between metal-binding sites
• discovering new binding motifs not represented in existing databases
• delivering a unified, chemistry-aware framework.

MDAP's data science specialists will co-design and train the machine learning model to integrate structural and experimental data on metal-binding sites. MDAP team will facilitate robust and reproducible pipelines that align with University standards for research excellence and open science. MDAP’s computational expertise will help build a transformative structural prediction resource to feasibly integrate artificial intelligence, structural biology, and genomics.

The project will produce the first genome-wide maps of bacterial metalloproteins and accelerate antimicrobial resistance research by identifying new metal-binding protein and revealing how metal ions sustain bacterial growth and virulence. Ultimately, findings will enable exploration of novel therapeutic targets and antimicrobial agents that disrupt bacterial activity.