On Jan. 13, 2026, collaborating with Google’s top artificial intelligence labs, a team of Yale researchers announced they have developed a large-language model that identified a new treatment for cancer that proved successful when tested on human skin and pulmonary cells.

When an artificial intelligence model suggested using a drug called Silmitasertib to help the human body locate cancerous tumors, Yale School of Medicine Professor David van Dijk said it was “surprising.” The model wasn’t relying on published literature it ingested in its training. In fact, no research had ever suggested that Silmitasertib could fight cancer in this way. 

But his lab’s custom AI model, called Cell2Sentence, hypothesized that the drug could increase antigen presentation — a process that helps the body’s immune system identify rogue cells, which in turn helps the body fight cancer. Soon enough, his lab at the Yale School of Medicine tested the hypothesis on human skin and pulmonary cells. His lab’s AI model was right. The van Dijk lab published its findings in a preprint paper in October in collaboration with Google’s top AI labs, Google DeepMind and Google Research. 

Researchers in Yale’s van Dijk lab have been working for years on a new approach to understanding the human cell. By analyzing single-cell RNA sequences, which show which genes are being expressed in each individual cell, the lab aims to understand the human body at a cellular level of detail.

The researchers hope that their findings can help shortcut this pre-clinical phase, with AI models “guiding researchers to focus their laboratory efforts on experiments with the highest probability of success,” according to Azizi.