Published in Science: IBM and Memorial Sloan Kettering researchers solve cancer immunotherapy mystery

Share this post:

Cancer immunotherapies, treatment approaches which harness a person’s own immune system to target and kill cancer cells, are currently a major driver in the development of new cancer treatments.  With the advent of next-generation gene editing technologies like CRISPR-Cas9, one could imagine a future in which cancer treatments are tailor-made not only for particular cancer subtypes, but also for every individual patient. These personalized medicines could equip a patient’s immune cells with weapons that uniquely identify and destroy mutant cancer cells, while leaving the body’s normal, healthy cells unharmed.


Figure 1: Cancer peptide (yellow) bound to an HLA protein (blue), a human protein molecule responsible for immune cell recognition of cancer. New work associates cancer peptide bridging structures in the HLA (red) with poor immunotherapy outcomes

Realizing this future, however, depends on getting a grasp on the molecular interactions that allow current immunotherapies to be effective.  The most transformative immunotherapies of the past decade, called immune checkpoint inhibitors, boost the body’s fight against cancer cells by switching off the systems that normally restrain the teeth of our immune system, our “killer” T cells. Effectively activating our very own immune system has produced significant results in some patients suffering from late-stage skin, lung, and other cancers, helping immune cells identify and neutralize cancer cells that had spread throughout patients’ bodies.1

Unfortunately, these therapies are currently only effective in about 10-30 percent of patients to whom they are prescribed.  Science is faced with the challenge: how can immunotherapy effectiveness be extended to all cancer patients?

In research recently published in Science, scientists at IBM, Columbia University and the Memorial Sloan Kettering Cancer Center (MSKCC) have solved an important piece of the immunotherapy puzzle. MSKCC researchers discovered that genes associated with how killer T cells recognize cancer cells are essential for immunotherapies’ success.  Specifically, these genes produce human leukocyte antigens (HLAs), protein molecules that bind to cancer-specific peptides and appear on the cell’s surface. The presence of these HLAs allows killer T cells to identify and destroy their cancer cell targets. These scientists found that patients with certain genes that result in specific types of HLA molecules, such as the HLA-B44 supertype, responded much better to immunotherapies than average, while patients with other particular HLA genes did not respond as well. They also showed that patients harboring tumors with very high mutation rates responded disproportionately well to these immune checkpoint inhibitor treatments. 2

Why do specific HLA proteins lead to drastically different immunotherapy outcomes?  To answer this question, scientists at IBM Research conducted large scale computer simulations and machine learning techniques on these different HLA molecules in atomistic detail.2 Remarkably, the scientists discovered that several HLA proteins associated with poor therapeutic outcomes had structural appendages (HLA bridges) that closed over the bound cancer peptide. The team theorized that these structures likely prevent the killer T cells from recognizing the cancer cells. In two separate studies of patients who only had moderate response to the immunotherapies, these HLA bridge structures were somewhat flexible, opening and closing in motions that still likely provide some access to T cell receptors.  In the worst performing group of patients, the HLA bridge proved to be very rigidly closed, likely preventing any significant binding with and recognition by the cancer-killing T cells. A technical explanation of these structures can be found in Figure 2.

Connecting these atomically detailed computer simulations with clinical data derived from thousands of patients is unprecedented, and the results could have immediate implications for clinicians when treating cancer. Specifically, this study could help to inform clinicians on whether they should prescribe existing immunotherapies to patients with certain genes, and provides a framework for evaluating whether future immunotherapies will be well complemented by particular patient genotypes.  More distantly, and requiring further investigation and research, the IBM and MSKCC scientists’ work might help suggest strategies for developing new HLA-based cancer immunotherapies, and highlights important facets of protein-protein interactions that may form the basis for personalized medicines of the future.

Read more about new research from the IBM Healthcare & Life Sciences team on our website



Figure 2 (left). Molecular dynamics simulations of HLA-B*15:01 (one of the representative weaker responders). (A) Overview of the three-dimensional structure of the peptide-binding groove of HLA-B*15:01, light purple; bound epitope, yellow; bridging residues, light pink. (B) Side view of the bridge-sequestration effect over bound-peptide residue positions P2 and P3 (light blue and red, respectively). (C) MD simulation snapshots of both the isolated HLA B*15:01 molecule and its complex with a 9-mer epitope; each trajectory was run over the course of 500 ns of simulation time. (D) Observables from the MD simulations described in (C). Shown is the mean bridge distance in the HLA-B*15:01 molecule and in the HLA-B*15:01-epitope complex. The residue-position root mean square fluctuations (RMSFs) indicate that each of the bridging residues becomes more rigid in the presence of the epitope.






1) T. Chan and coworkers, New England Journal of Medicine 2014, 371, 2189-2199.





More Healthcare stories

Dark Matter Matters: AI Makes DNA Dark Matter Useful

What is the minimal description that captures a space? Asking a mathematician’s basic question of a  biological dataset reveals interesting answers about biology itself. This summarizes our underlying approach to subtyping hematological cancer. Disease subtyping is a central tenet of precision medicine, and is the challenging task of identifying and classifying patients with similar presentations […]

Continue reading

Helping to Untangle Cancer Drug Resistance with Data

Why do targeted cancer therapies often fail? We have acquired so much more understanding about cancer in the last fifty years than in the last five thousand years. Approaches to patient treatments have dramatically changed, and statistics show significant improvement in patient response and outcomes to therapy in the last half a century [1]. Yet […]

Continue reading

Novel AI tools to accelerate cancer research

At the 18th European Conference on Computational Biology and the 27th Conference on Intelligent Systems for Molecular Biology, IBM will present significant, novel research that led to the implementation of three machine learning solutions aimed at accelerating and guiding cancer research.

Continue reading