Columbia University Researchers Develop AI to Predict Gene Activity and Cell Function
Researchers at Columbia University’s Vagelos School of Medicine have developed a new artificial intelligence (AI) system called the “Universal Expression Transformer” (GET). This model is capable of accurately predicting gene expression in any human cell. This discovery, which was recently published in the prestigious Nature journal, can completely change the way cancer and other genetic diseases are studied.
AI Changed the Approach to Biology
Raul Rabadan, lead author of the research and professor of systems biology, said, “This new model is an important step towards turning biology into a predictive science. Traditional biological research methods are only able to explain how cells perform their functionality, but they cannot tell how cells will function in the future or how they will respond to changes.”
According to Rabadan, “With this AI model, scientists can now quickly and accurately understand biological processes inside cells.” This method provides the ability to conduct large-scale research by complementing traditional laboratory techniques.
How Does the GET Model Work?
In this research, Rabadan and his team trained a machine learning model using gene expression data from millions of human cells. The model included genome sequences as input and data on which parts of the genome are active and expressed.
This model works similar to popular AI systems such as ChatGPT. Just as language models understand the “grammar” and apply it to new contexts, this AI model learns the “grammar” of different cellular states and applies it to predict diseased or normal cells.
Guide to New Discoveries
The GET model was trained on data from more than 1.3 million human cells and is able to accurately predict cell types it has never seen before.
In addition, the team also demonstrated that this AI system can reveal the hidden biological mechanisms of diseased cells. Using the example of a hereditary childhood leukemia, the researchers showed how changes in genes interfere with the action of two different transcription factors. The AI’s predictions were also found to be correct in laboratory tests.
Discovery of “Dark Matter”
This AI system also provides an opportunity to study the “dark matter” of the genome. This part of the genome does not contain protein coding genes and has been little researched so far. Rabadan said, “Most mutations in cancer patients are found in this dark region. With the help of these models, the functionality of these regions can now be revealed.”
New Hopes for the Future
Rabadan’s team is currently researching several types of cancer, including brain and blood cancer. Their goal is to understand the regulatory “grammar” in normal cells and see how cells change during cancer development.
This discovery could help understand many diseases other than cancer and find new treatments for them. Scientists can now insert new mutations into computer models to understand how they affect cells.
This research is a great example of the growing use of AI in biology. Rabadan said, “We are on the threshold of turning biology into a predictive science. This is an exciting era.”
This study will not only help in understanding complex diseases like cancer, but it will also open doors to new possibilities in the medical field.
