Google и Йельский университет: Новый ИИ для борьбы с раком обнаруживает горячие опухоли Google and Yale University: New AI for Cancer Treatment Identifies Hot Tumors

The Google corporation, in collaboration with Yale University, has unveiled a new foundational model consisting of 27 billion parameters, designed to interpret the «language» of individual cells.

C2S-Scale 27B generated a hypothesis regarding the behavior of cancerous cells, which was later confirmed through experiments conducted on live organic samples.

«This discovery has opened up a promising avenue for the development of new cancer therapy methods,» the company emphasized.

The model is built upon previous research that demonstrated how biological and linguistic systems adhere to similar scaling laws—efficiency increases with size.

A significant challenge in cancer immunotherapy is that many tumors remain «cold,» rendering them invisible to the immune system. One method to «warm» them up is to induce the presentation of signals through a process called «antigen presentation.»

Google tasked C2S-Scale 27B with finding a compound that functions as a conditional enhancer: it boosts the immune response only in a specific «immunopositive» environment where a low level of interferon is present, but insufficient for independent activation of antigen presentation.

This task required dealing with conditional reasoning, which smaller models struggled to manage.

To achieve this goal, researchers developed a virtual two-context screening capable of revealing this synergistic effect. It involved two phases:

Subsequently, Google modeled over 4,000 compounds in both contexts and instructed the model to determine which of them enhanced antigen presentation solely in the first context. This allowed the team to focus the search on clinically relevant scenarios.

Among numerous options, 10–30% had already been mentioned in scientific literature, while the remainder turned out to be unexpected discoveries.

The model identified a «striking contextual gap» for the CK2 kinase inhibitor named silmitasertib (CX-4945). The neural network predicted a significant enhancement in antigen presentation when using the drug in an «immunopositive» context, but nearly no effect in an «immuno-neutral» context.

Notably, this represents a completely new idea that had not been mentioned previously.

In the subsequent phase, researchers tested the hypothesis in the lab. They utilized human neuroendocrine cells—samples that the model had not «seen» during training. The results indicated that:

In laboratory experiments, this combination led to approximately a 50% increase in antigen presentation, making tumors more visible to the immune system.

The digital prediction was repeatedly validated.

C2S-Scale discovered a new conditional interferon amplifier that could aid in transforming «cold» tumors into «hot» ones—more responsive to immunotherapy.

«While this is merely the first step, it already provides an experimentally validated foundation for developing new combination therapies where multiple drugs work together for a stronger effect,» the blog stated.

Yale University teams are already investigating the identified mechanism and testing other AI predictions in various immune contexts. With further preclinical and clinical validation, such hypotheses could expedite the development of new treatment methods.

Previously, the biotechnology company SpotitEarly began developing a home test for cancer based on analyzing human breath. The technology combines the olfactory abilities of dogs with artificial intelligence algorithms.

As a reminder, in September, scientists developed an AI tool for predicting over 1,000 diseases and forecasting health status changes over the next 10 years.