Google DeepMind
Google’s DeepMind and
Yale University have joined forces to create a revolutionary AI model that
could change the future of cancer research. The new model, named C2S-Scale
27B, has helped scientists uncover how cancer cells hide from the immune
system — and how to expose them. A Breakthrough in Cancer Science According to DeepMind,
this 27-billion-parameter foundation model is built on Google’s Gemma open
models and is specially “designed to understand the language of individual
cells.” In a blog post, Google
described this as a milestone moment for AI in science, saying the model
“generated a novel hypothesis about cancer cell behaviour, which scientists
later validated in living cells.” This achievement builds on DeepMind’s earlier
work showing how large AI models can demonstrate advanced reasoning capabilities
— beyond just data analysis. How Does C2S-Scale 27B Work? One of the biggest
challenges in cancer immunotherapy is that tumours often remain
“invisible” to the immune system. To fight this, researchers try to make these
tumours display “immune-triggering signals,” a process known as antigen
presentation. DeepMind’s AI was
trained to find a drug that acts as a conditional amplifier — a compound
that helps immune cells recognise and attack cancer cells under specific
conditions. To achieve this, C2S-Scale
27B used what researchers call a “dual-context virtual screen.” The
model analysed over 4,000 drugs across different tumour samples and
isolated cell data. This large-scale simulation helped identify which compounds
could enhance immune activation in particular biological contexts. Discoveries Beyond Expectations Interestingly, while
some of the drugs tested were already known to target cancer cells, the AI also
identified several previously unknown compounds with potential
cancer-fighting properties. These predictions were
later tested in the lab using human neuroendocrine cell models — a type
of cell that wasn’t even part of the model’s training data. Among the tested
drugs, one called CX-4945 stood out, with results showing it could boost
antigen presentation by nearly 50 per cent. Why This Matters The discovery
highlights that C2S-Scale 27B didn’t just process data — it reasoned through
complex biological interactions to predict how specific conditions could
influence treatment outcomes. This marks a huge
shift in how cancer research can be done. Instead of relying solely on years of
trial-and-error testing, scientists can now simulate drug reactions
virtually, accelerating discoveries and reducing costs dramatically. Google’s breakthrough
shows how artificial intelligence is moving beyond language and image
generation — and stepping into the frontlines of medical innovation.
Google says its AI model could open new cancer drug pathways. (Image Source: Google)
’s New AI Model Cracks a Major Cancer Code, Paving the
Way for Next-Gen Therapies
Tags
BRAKING NEWS