A new artificial intelligence tool called Corgi can predict how genes behave across different cell types by considering the unique molecular environment inside each cell, potentially revolutionizing how scientists model disease and discover new drugs. Developed by Martin Vingron's bioinformatics group at the Max Planck Institute for Molecular Genetics in Berlin, the tool addresses a fundamental limitation in earlier AI models that ignored cellular context when predicting gene activity.

Why Does Cellular Context Matter for Gene Prediction?

Every cell in your body contains the same DNA, yet a neuron behaves completely differently from a liver cell. The reason lies not just in the genetic code itself, but in the regulatory molecules present in each cell type. Earlier AI models failed to account for this reality, treating DNA as if it operated in isolation. "The same DNA sequence may behave differently in a neuron than in a liver cell, for example, because each cell contains a different mix of regulatory molecules," explained Ekin Deniz Aksu, the study's first author.

This oversight meant that older models couldn't reliably predict gene behavior beyond their training data. Corgi changes this by integrating expression data from a large set of regulatory genes, creating an architecture that more closely mirrors actual biological reality inside a cell.

What Can Corgi Actually Predict?

Corgi demonstrates impressive accuracy across multiple measures of genome activity. The model can correctly predict measurements related to how genes are expressed, including gene expression levels, chromatin accessibility (how tightly DNA is packaged), and histone modifications (chemical tags on DNA that control gene activity). An improved version called Corgi+ can even infer epigenetic information from RNA sequencing data, filling gaps in datasets without requiring researchers to conduct every experiment directly.

This capability opens doors for studying rare cell types, embryonic cells, and limited patient tissue samples, where experimental data is scarce or difficult to obtain.

How Could This Tool Transform Drug Discovery and Disease Research?

The implications extend far beyond basic research. Corgi could help merge sequence-to-function models with virtual cell systems, creating digital representations of cells that incorporate DNA activity information. These enhanced virtual cells could enable scientists to screen drug candidates computationally before testing them in the lab, potentially accelerating development timelines and reducing costs.

Understanding how genetic variants alter cellular processes is particularly important for personalized medicine. Although each person carries small genetic variations, only a fraction are functionally relevant to disease. Researchers have increasingly focused on these variants because they may explain disease mechanisms and could inform tailored medical interventions.

Steps to Leverage AI for Genomic Research

• Integrate Cellular Context: Use AI models that account for the unique regulatory environment of different cell types rather than treating DNA sequences in isolation, improving prediction accuracy across diverse cell populations.
• Combine Multiple Data Types: Leverage expression data from regulatory genes alongside sequence information to train more robust models that reflect biological reality and generalize better to new scenarios.
• Fill Experimental Gaps: Deploy AI tools to infer missing epigenetic and expression data for rare cell types or limited tissue samples, reducing the need for expensive or difficult-to-obtain experimental measurements.

"With Corgi, we have created a model that overcomes many of these limitations. It can correctly predict measurements related to genome activity, including gene expression, chromatin accessibility, and histone modifications," said Ekin Deniz Aksu.

Ekin Deniz Aksu, Study First Author, Max Planck Institute for Molecular Genetics

The research represents a significant step forward in computational biology. By respecting the complexity of cellular environments, Corgi demonstrates how AI can move beyond simplified assumptions to capture the nuances of real biological systems. As researchers continue refining the tool, it could become instrumental in understanding how mutations cause disease and in identifying promising drug targets more efficiently than traditional methods.

The findings were published in Nature Communications and reflect growing recognition that context matters in genomics. Future applications could include virtual screening of therapeutic compounds, modeling how genetic variants affect cellular behavior in specific tissues, and ultimately supporting the development of more effective, personalized treatments.