AI Model Could Speed Tuberculosis Vaccine Research

Researchers in Phoenix report that the approach predicted immune recognition of tuberculosis antigens from genetic sequence alone.

Updated on Nov. 4, 2026 in Diseases

Sunlit research laboratory bench with a sealed tray of clear sample vials beside a translucent sculptural model of interlocking helices.
Researchers at the Translational Genomics Research Institute in Phoenix report an AI model predicted tuberculosis antigen recognition by immune cells from genetic sequence alone. AI Illustration. Upload story photo >

Researchers at the Translational Genomics Research Institute (TGen) in Phoenix used AlphaFold3, a structure-prediction model introduced in 2024, to predict whether immune T cell receptors recognize specific antigens from genetic sequence alone. The findings were published in Science Advances, and they could matter most for people with tuberculosis and for future vaccine and immunotherapy research.

Why it matters

Predicting whether a T cell receptor recognizes a given antigen has been a long-standing challenge in immunology, and a reliable shortcut would speed the search for targets in infection and cancer. Tuberculosis made a demanding test because the bacterium carries many antigens while the immune molecules that present them vary widely between people and populations. New tuberculosis vaccines and diagnostics are urgently needed, which is why the work is being watched closely.

This is a computational study in which AlphaFold3 was first tested on immune complexes with experimentally determined structures, then applied to eight MHC class II-antigen pairs from the tuberculosis-causing bacterium. The model was able to predict T cell receptor recognition of antigens known only from genetic sequences, though the reported sample is small and the approach remains unvalidated at scale.

The players

Translational Genomics Research Institute (TGen)

A Phoenix-based research institute focused on genomic medicine, including work on cancer and infectious disease, that conducted this study.

AlphaFold3

A structure-prediction model introduced in 2024 that was designed to predict interactions among multiple molecules, used here to model immune complexes.

The details

AlphaFold3 models how an MHC molecule, a bound peptide fragment of an antigen, and a T cell receptor fit together in three dimensions. Because the shape of that complex largely determines whether the T cell responds, an accurate structural model can serve as a proxy for recognition. The TGen team is also building graph-based neural networks, trained on data from hundreds of thousands of peptide interactions, to predict how strongly MHC class II molecules bind peptides.

Timeline

  1. AlphaFold3 was introduced in 2024.

  2. The TGen T cell recognition research was reported on November 3, 2026.

Health Landscape

Predicting T cell recognition has been a persistent bottleneck in immunology, and this work applies the AlphaFold3 structural framework to that problem in tuberculosis. If validated more broadly, the approach could complement experimental methods that identify which antigens immune cells target.

For most readers this is a research development rather than a change to care, so there is nothing to act on today. People with a history of tuberculosis exposure or immune-related conditions may want to ask their physician about current testing and vaccine options as new tools emerge.

The takeaway

Computational tools that predict immune recognition could eventually speed the search for tuberculosis vaccines and cancer immunotherapy targets, though the approach is still under validation. Anyone with questions about tuberculosis testing or immune health should raise them with their physician.

Further reading

For more on how immune responses are studied and treated, explore our Allergies coverage.