Researchers Harness Artificial Intelligence to Decode Gene Networks and Uncover Disease Mechanisms

Christina Theodoris and her colleagues at Gladstone Institutes, the Broad Institute of MIT and Harvard, and Dana-Farber Cancer Institute trained a computer model to understand how genes interact. Photo: Michael Short/Gladstone Institutes
Christina Theodoris and her colleagues at Gladstone Institutes, the Broad Institute of MIT and Harvard, and Dana-Farber Cancer Institute trained a computer model to understand how genes interact. Photo: Michael Short/Gladstone Institutes

In a groundbreaking study, researchers from Gladstone Institutes, the Broad Institute of MIT and Harvard, and Dana-Farber Cancer Institute have utilized artificial intelligence (AI) to gain a deeper understanding of the intricate connections between thousands of human genes and how disruptions in these networks contribute to the development of diseases.

The team, led by Gladstone Assistant Investigator Christina Theodoris, MD, PhD, developed a cutting-edge AI model called Geneformer. By training Geneformer on vast amounts of data on gene interactions from various human tissues, the researchers were able to predict how these networks malfunction in the context of different diseases.

The findings, published in the journal Nature, have significant implications for the field of biology and could pave the way for targeted therapeutic interventions in diseases with limited treatment options.

Gene Networks: Complex Web of Connections

The human genome comprises around 20,000 genes that work together in intricate networks to regulate cellular function. When certain genes are activated, they trigger a cascade of molecular events that influence the activity of other genes, creating a web-like structure. Understanding these gene networks and how they change in the presence of disease is a daunting challenge.

Theodoris and her team addressed this challenge by employing transfer learning, a machine learning technique, to train Geneformer. Transfer learning involves pretraining the AI model on a large dataset to acquire foundational knowledge of gene interactions, which can then be applied to various tasks without requiring extensive retraining.

By fine-tuning Geneformer to analyze gene connections and predict disease outcomes, the researchers achieved remarkable accuracy even with limited data. This breakthrough could have significant implications for rare diseases and those affecting tissues that are challenging to sample in clinical settings.

Unraveling Heart Disease

To demonstrate the power of Geneformer, the researchers focused on heart disease. Geneformer successfully identified key genes associated with heart disease, reaffirming existing knowledge while also uncovering previously unknown genes like TEAD4, which plays a crucial role in the function of heart muscle cells.

Moreover, Geneformer identified potential gene targets that could restore diseased heart cells to a healthy state. Experimental validation using CRISPR gene editing technology confirmed the effectiveness of these targets in restoring the normal beating ability of diseased heart muscle cells.

Transfer Learning as a Game-Changer

One of the advantages of using Geneformer is its ability to predict which genes can be targeted to switch cells from a diseased to a healthy state. Unlike traditional approaches that require retraining models from scratch for each new application, the transfer learning approach allows Geneformer’s foundational knowledge to be applied to a wide range of biological questions.

The researchers aim to expand the scope of Geneformer by analyzing additional cell types and making the model open-source, enabling other scientists to leverage its capabilities.

The groundbreaking study, titled “Transfer learning enables predictions in network biology,” opens up new avenues for understanding gene networks and devising targeted therapies for various diseases. With the integration of AI and genetics, the possibilities for unlocking the secrets of human biology are limitless.

Explore more topics
Author
Prolific News
Related news stories
Advertisement