Scientists have developed a groundbreaking new artificial intelligence (AI) model called GET (General Expression Transformer) that predicts gene expression with remarkable accuracy. This model significantly enhances our understanding of gene regulation and has the potential to revolutionize fields like medicine and biotechnology. GET uses a novel approach, learning the rules of transcriptional regulation directly from chromatin accessibility data.
How GET Works: A Sneak Peek into the Cellular Machinery
GET analyses massive datasets of chromatin accessibility and gene expression across numerous human cell types. Specifically, it focuses on the accessibility of DNA regions (how easily proteins can bind to them) and the presence of transcription factors—proteins that control gene activity. This information is then processed by a sophisticated machine learning algorithm to predict gene expression. Think of it as a complex decoder ring for the instructions embedded within our DNA.
Self-Supervised Learning: Teaching the AI to Learn
Interestingly, GET applies a technique called self-supervised learning. Researchers mask parts of the input data and train the AI to predict the missing information. This forces the model to learn deeper relationships and patterns within the data, resulting in a more robust and accurate predictor of gene expression.
Predicting Gene Expression in Unseen Cell Types
One of GET’s most impressive features is its ability to accurately predict gene expression in cell types it hasn’t explicitly trained on – a feat that previous models rarely achieve. This generalizability is critical for studying a wide range of biological processes and diseases.
Beyond Prediction: Uncovering Hidden Regulatory Mechanisms
GET doesn’t just predict; it also provides valuable insights into the complex interactions between different components of the gene regulatory machinery. By analysing the model’s predictions, researchers can identify cis-regulatory elements (DNA sequences that control gene expression), upstream regulators, and even pinpoint potential interactions between different transcription factors. This level of detail offers new avenues for research into various diseases, including cancers and genetic disorders.
Identifying Novel Regulatory Interactions
Furthermore, the GET model has already revealed previously unknown regulatory interactions. For instance, it identified a lymphocyte-specific interaction involving PAX5 and nuclear receptor family TFs, potentially shedding light on the mechanisms driving specific diseases. This demonstrates the powerful potential of GET for uncovering hidden cellular mechanisms.
Zero-Shot Prediction: Predicting Regulatory Activity Without Training
Amazingly, GET can even perform zero-shot prediction of regulatory activity. This means it can predict the activity of DNA sequences in a specific cell type without needing prior training data for that cell type. This is a monumental leap forward in our ability to understand gene regulation across a wide range of conditions.
The development of GET represents a significant advancement in our understanding of gene expression. Certainly, it can predict gene expression in cell types that haven’t been studied yet. Additionally, it can also identify new interactions that regulate processes in cells. “Zero-shot predictions” mean making predictions without having seen similar examples before. So, these abilities create exciting new possibilities. Moreover, researchers can explore these in various fields and applications.
References
Fu, X., Mo, S., Buendia, A., Laurent, A. P., Shao, A., Del Mar Alvarez-Torres, M., Yu, T., Tan, J., Su, J., Sagatelian, R., Ferrando, A. A., Ciccia, A., Lan, Y., Owens, D. M., Palomero, T., Xing, E. P., & Rabadan, R. (2025). A foundation model of transcription across human cell types. Nature. https://doi.org/10.1038/s41586-024-08391-z
To stay updated with the latest developments in STEM research, visit ENTECH Online. This is our digital magazine for science, technology, engineering, and mathematics.
At ENTECH Online, you’ll find a wealth of information. Besides this, we offer insights and resources to fuel your curiosity. Specifically, our goal is to inspire your passion for new scientific discoveries.
Disclaimer
This article/blog post is not intended to provide professional or technical or medical advice. Please consult a healthcare professional before making any changes to your diet or lifestyle. AI-generated images are used only for illustration and decoration. Their accuracy, quality, and appropriateness can differ. Users should avoid making decisions or assumptions based only on the text and images.
