A team from the Vaglos School of Medicine at Columbia University has developed an innovative artificial intelligence (AI) model called the Universal Expression Transformer (GET), which can accurately predict gene activity within human cells and provide a new perspective for understanding the internal workings of cells. This breakthrough achievement can help scientists explore a range of health issues, from cancer to genetic diseases, in an unprecedented way, taking a big step forward in medical research. The relevant paper was published in the latest issue of the journal Nature. Traditional biological methods excel at describing the working principles of cells and how they respond to external changes, but lack the ability to predict cell behavior and its response to changes such as carcinogenic mutations. In contrast, the GET model is able to accurately predict cell activity, marking a shift in biology from a field that primarily relies on descriptive analysis to a science that can predict and regulate the underlying systems of cell behavior. This time, the team used AI to predict the active genes within specific cells, which is crucial for determining cell identity and function. They trained a machine learning model using gene expression data from millions of cells in normal human tissue, which includes not only genome sequences but also information about which parts of the genome are accessible and expressed. The overall idea of the GET model is similar to that of large language models such as ChatGPT: recognizing basic rules (such as language syntax) through training data, and then applying these rules to new scenarios. After data training, the GET model became accurate enough to predict gene expression patterns in unseen cell types, and the results were highly consistent with experimental data. In addition, the team also used the GET model to reveal the hidden biological mechanisms in diseased cells. In a specific case, a study on a hereditary childhood leukemia showed that AI successfully predicted that certain mutations would disrupt the interaction between two transcription factors that determine the fate of leukemia cells, and the experiment confirmed AI's prediction. This enhances people's understanding of the driving mechanisms of this disease. This study not only opens up new avenues for exploring the pathology of various diseases, but also provides the possibility for discovering new therapeutic targets. Now, scientists can understand and predict the specific effects of newly discovered mutations on cells by inputting them into computer models. (New Society)