Thanks to one of its AIs, Google may have found a new way to treat cancer

“A milestone for AI in science.” This is how Google announced the discovery of a new way to treat cancer thanks to one of its artificial intelligence models, Cell2Sentence-Scale 27B. Launched in collaboration with Yale University, it is designed to understand the behavior of individual cells.

As part of this collaborative research, the model generated a new hypothesis about the behavior of cancer cells. As Google explained in a blog post, one of the big difficulties with cancer immunotherapy is that many tumors are “cold.” In other words, they are invisible to the body’s immune system.

A key strategy to warm them up is to force them to emit signals that will trigger the immune response through a process called antigen presentation (MHC-I), which helps identify abnormal cells.

“Conditional amplifier”

Google and Yale University have thus entrusted a mission to the AI ​​model: “to find a drug that acts as a conditional amplifier, capable of reinforcing the immune signal.” But only in a specific “immunological context-friendly” environment, that is, in which low levels of interferon, a key protein in immune signaling, were already present, but not in sufficient quantities to induce antigen presentation on their own. The antigen is a substance that, once detected, triggers a reaction in the body’s defense system.

From there, the two partners designed a dual-context virtual display, with a “positive immune context” on one side and a “neutral immune context” on the other. For the first, the AI ​​model received samples from real patients with intact tumor immune interactions and low-level interferon signaling. For the second, data from isolated cell lines without immunological context were presented.

A potentially promising avenue

Google and Yale University then simulated the effect of more than 4,000 drugs in these two environments and asked Cell2Sentence-Scale 27B to predict which ones would only stimulate antigen presentation in the first. Thus, the model predicted that one molecule, silmitasertib, would likely strengthen the immune system’s ability to recognize cancer cells or show a strong increase in antigen presentation.

Although this CK2 protein kinase molecule had already been studied in other medical contexts, until now it was not known that it was capable of improving the visibility of tumors for the immune system. And this breakthrough prediction was confirmed in laboratory tests, where the combination of silmitasertib and low-dose interferon resulted in an approximately 50% increase in antigen presentation.

“While this is a first step, it offers a promising and experimentally validated avenue for the development of new combination therapies, using multiple drugs together to achieve a more potent effect,” the company added.