RNA modifications are like fine-tuning knobs for how our genes are expressed. One such modification, N4-acetylcytidine (ac4C), plays a crucial role in keeping RNA stable and functional—and it’s been linked to several human diseases. But finding where these ac4C modifications occur in RNA molecules isn’t easy.

Researchers are turning to artificial intelligence to solve this problem. Researchers from Xidian University presents a powerful new tool called TransCNN, which combines two AI models—transformers and TextCNN—to accurately predict where ac4C modifications happen in human RNA sequences.

Why is this important? By knowing where ac4C sites are, scientists can better understand how gene expression is regulated and how RNA functions in health and disease. But older prediction models weren’t very accurate, largely because they struggled to understand the context of RNA sequences.

TransCNN improves on this by blending the transformer model, which is great at capturing complex relationships in sequences, with TextCNN, which detects patterns at different scales—both close-up and wide-range. In tests, TransCNN outperformed previous methods with an impressive accuracy of over 83%.

This AI-driven approach opens the door to better understanding of RNA biology and could eventually lead to better diagnostics or treatments for diseases where RNA regulation goes wrong.

The datasets and code used in this study are available at https://github.com/liukai23157/TransCNN.