AI Model Predicts Gene Activity in Human Cells, Transforming Biological Research

AI Model Revolutionizes Gene Activity Prediction

Scientists at Columbia University have developed a groundbreaking artificial intelligence (AI) model that can accurately predict gene activity in human cells, potentially transforming biological research and our understanding of diseases. The model, named General Expression Transformer (GET), was trained on data from over 1.3 million normal human cells, spanning 213 different cell types 1

1

2

3

How GET Works

GET uses an approach similar to language models like ChatGPT, learning the "grammar" of gene regulation. By analyzing genome sequences and data on which parts of the genome are accessible and expressed, GET can predict which genes will be active in specific cell types, even those it hasn't encountered before 1

1

2

3

Implications for Cancer Research

The AI model has already shown promise in uncovering mechanisms driving diseases. In one application, GET helped researchers understand the underlying causes of an inherited form of pediatric leukemia. The model predicted that specific mutations disrupt the interaction between two transcription factors that determine the fate of leukemic cells, a finding later confirmed by laboratory experiments 2

2

3

4

Exploring Genomic "Dark Matter"

GET could also help scientists explore the genome's "dark matter" – regions that don't encode known genes but where most cancer-related mutations occur. This could lead to new insights into cancer development and potential treatment targets 2

2

3

4

Transforming Biology into a Predictive Science

The development of GET represents a significant step towards turning biology into a more predictive science. It allows researchers to conduct large-scale computational experiments, potentially reducing the need for time-consuming and costly laboratory work 2

2

3

4

Related Stories

AI Designs DNA to Control Gene Expression in Healthy Mammalian Cells

AI Breakthrough: Designing Synthetic DNA Switches for Precise Gene Control

Scientists create AI model that maps how genes work together inside human cells

Google DeepMind's AlphaGenome AI tackles the 98% of human genome that doesn't code for proteins

Comparison to Other AI Breakthroughs in Biology

GET's potential impact is being compared to that of AlphaFold, the AI system that predicts protein structures and was recognized with the 2024 Nobel Prize in Chemistry. While AlphaFold focuses on protein structure, GET addresses the equally fundamental question of gene regulation 1

1

5

Future Applications

Researchers are already using GET to study various cancers, from brain to blood cancers, and to understand how cells change during cancer development. The model could also aid in the development of cell-specific gene therapies and help scientists prioritize which experiments to conduct in the lab 1

1

2

3

4

As AI continues to make inroads in biology, tools like GET are poised to accelerate scientific discovery and deepen our understanding of cellular processes. This could lead to breakthroughs in treating not only cancer but a wide range of genetic diseases, ushering in a new era of predictive biology 1

1

2

3

4

5