Real-Time Pathogen-Host Interactome Prediction

The contract seeks to develop a computational system capable of rapidly characterizing novel or emerging pathogens—viruses, bacteria, and parasites—using only their protein sequence data, eliminating the need for weeks or months of experimental work. The system must accurately predict protein-protein interactions between pathogens and human hosts across all three pathogen classes, with zero-shot learning capabilities to generalize to previously unseen threats without prior exposure in training data. It must provide detailed functional annotations for both pathogen and host proteins, automatically generate ranked mechanistic hypotheses about infection pathways, and deliver core predictions within 15 minutes and full characterization reports within one hour using standard computing hardware. Rigorous validation is required to ensure the system does not simply memorize known data but can truly generalize, with performance benchmarks against established protein interaction databases and experimental confirmation via standard binding assays. The initiative, titled Real-Time Pathogen-Host Interactome Prediction under solicitation DPA26BZ03-DV014, is issued by the Defense Advanced Research Projects Agency within the Department of Defense to address critical delays in therapeutic response to biological threats. The effort is structured as a total small business set-aside, limited to entities with fewer than 500 employees, aligning with SBIR/STTR mandates. The system must be operationally deployable in real-world defense and public health scenarios to enable rapid prioritization of medical countermeasures and enhance force health protection. Proposals must demonstrate not only technical feasibility but also robust evaluation methodologies and the ability to operate effectively under standard computational environments, ensuring accessibility and scalability for urgent response operations.