Texas A&M University researchers have developed a suite of artificial intelligence tools designed to accelerate the traditionally slow pipeline for tuberculosis drug discovery. Led by Dr. James Sacchettini, alongside researchers Siddhant Rath and Saswati Panda, the team’s work addresses a critical bottleneck in combating a disease the World Health Organization ranks as the world’s deadliest infectious pathogen. Standard tuberculosis therapies require months to administer, with drug-resistant and HIV-co-infected cases demanding even longer treatments. Compounding the challenge, Mycobacterium tuberculosis possesses a thick, waxy cell envelope that blocks most pharmaceuticals, while its slow growth rate extends laboratory experiments from days to months. To streamline development, the researchers first established DAIKON, an open-source data management platform launched in 2023. Deployed across the Gates Foundation-funded Tuberculosis Drug Accelerator consortium, DAIKON centralizes fragmented research data, tracking a drug target from initial genetic analysis through years of chemical testing. Building on this infrastructure, the lab recently integrated two AI-driven systems to automate decision-making and data retrieval. The first tool, CAGE-Fusion, directly targets the high rate of false positives common in early-stage compound screening. During initial trials, thousands of molecules are tested against protein targets, but many produce misleading signals due to chemical clumping, assay interference, non-specific binding, or unintended reactive properties. These nuisance compounds waste months of research and hundreds of thousands of dollars. CAGE-Fusion analyzes historical screening data to classify problematic molecules across these four failure modes. Trained to distinguish active compounds from false signals, the model achieves approximately ninety-four percent accuracy in prioritizing suspicious candidates. When flags are raised, the system provides chemists with a breakdown of the specific molecular regions driving the false signal, enabling rapid triage before compounds enter costly late-stage development. A second AI system addresses knowledge fragmentation across the TBDA network. Historically, molecular history and experimental outcomes were scattered across network drives and presentation archives. The new platform ingests years of consortium data, converting structural and experimental records into a searchable knowledge base. Researchers can now trace a molecule’s developmental trajectory across multiple institutions, visualize stalled projects, and query results through an interactive chat interface. The system instantly retrieves associated presentations, metadata, and team evaluations, drastically reducing the time required to locate relevant historical data. Deployed amid a 2026 research landscape characterized by unprecedented computational capacity, these tools represent a strategic shift from AI-driven de novo design to intelligent workflow optimization. According to the researchers, the models are not intended to autonomously generate clinical candidates but to eliminate unviable paths early and aggregate institutional knowledge. By filtering false positives at scale and democratizing access to shared experimental records, the systems promise to compress the tuberculosis drug development timeline, reduce resource expenditure, and strengthen global efforts against a persistent pathogen.