Bioinformatics and AI modelling of genomic data has a new entrant that promises to make understanding complex gene regulation far simpler and faster. Corgi, presented in Nature Communications, is context-aware and can predict the activity of 16 different genomic assays, including RNA sequencing, chromatin accessibility and histone modifications.
What happened
Researchers have introduced Corgi (Context-aware Regulatory Genomics Inference), a model that mimics cellular gene regulation. It uses the expression of trans-regulators (transcription factors, coactivators, chromatin modifiers and RNA-binding proteins) as a context vector. Corgi's architecture combines convolutional and transformer layers, processing the input DNA sequence and integrating it with context data through FiLM layers. That lets the model capture both local cis-regulatory landscapes and long-range genomic interactions, such as enhancer-promoter loops.
Why it matters
Corgi performs strongly even on the hardest tasks, beating existing models such as EpiGePT. There is also an improved version, Corgi+, tuned for imputing epigenomic tracks in previously unseen cell types using RNA sequencing data. That opens new options for researchers, allowing more accurate prediction of genomic variant effects and identification of key cell-type-specific trans-regulators without prior training on those cell types.
What it means in practice
For developers and researchers working with AI in biology and medicine, Corgi is a powerful tool for:
- Accelerating research: more accurate and faster prediction of genomic activity cuts the time spent validating hypotheses experimentally.
- Developing new therapies: understanding gene regulation at this level could lead to more effective treatments for genetic diseases.
- Building personalised medicine: being able to predict variant effects in specific cell types brings individualised treatment approaches closer.
- Integrating data: the model effectively combines different types of genomic data, making analysis and interpretation easier.
- Advancing AI models: Corgi's architectural choices can be adapted to other problems that call for context-aware modelling of complex systems.