In a breakthrough that could reshape the future of medicine, scientists from Columbia University have developed an advanced AI tool designed to predict how genes inside cells influence their behavior.
This new tool, called the General Expression Transformer (GET), has the potential to open up new possibilities in treating genetic diseases, including cancer, by offering a deeper understanding of gene regulation.
AI Model 'GET' Unveils New Insights into Gene Expression for Targeted Therapies
GET works similarly to language models like ChatGPT, which learn grammar and sentence structure from vast amounts of text.
According to the Washington Post, instead of language, GET learns the "grammar" of genes, focusing on how they are turned on or off, or dialed up and down, to control the production of proteins.
These proteins are vital for nearly every process in the body, from moving and breathing to fighting diseases.
The model has been trained to understand gene expression — how genes function and interact within different cell types.
What makes GET unique is its ability to predict gene activity across various cell types. Unlike previous AI models that focused primarily on abnormal cells, such as cancer cells, GET was trained on data from more than 1.3 million normal human cells.
This broad dataset allows the model to make accurate predictions about gene expression even for cells it has never directly seen. For example, it could predict how genes behave in one type of cell based on what it learned from others.
The potential implications of this technology are vast. Gene expression plays a critical role in understanding how diseases arise, including genetic disorders.
By predicting how mutations affect specific cell types, researchers hope to develop more targeted gene therapies — treatments that focus on correcting genetic mutations in one specific type of cell without impacting others.
This precision could be a game changer for conditions like cancer, where thousands of mutations might be present but only a few are responsible for the disease's progression.
GET AI Model Predicts Cellular Behavior, Offering Hope for Leukemia
The success of GET is also tied to its ability to predict cellular behaviors in response to changes or disruptions, such as those caused by cancer-causing mutations.
In one experiment, the model successfully predicted how mutations in a gene associated with pediatric leukemia affected the behavior of leukemic cells, NeuroScience reported.
This prediction was later confirmed by laboratory experiments, underscoring the model's accuracy and usefulness in understanding complex diseases.
In addition to revealing hidden mechanisms behind diseases, GET could also help scientists explore the so-called "dark matter" of the genome — the vast regions of DNA that do not code for known genes but are thought to contain important regulatory information.
By analyzing these areas, researchers might uncover hidden mutations that contribute to conditions like cancer, opening new paths for therapies.
This AI-based approach is still in its early stages, but its potential is immense. Experts believe that much like AlphaFold2 revolutionized our understanding of protein structures, GET could pave the way for new, more effective treatments by improving our ability to predict and manipulate gene activity.
As AI continues to evolve, it promises to turn biology into a more predictive science, where scientists can not only observe but forecast the behavior of genes and cells in health and disease.
