Mapping out how genes are controlled inside cells is one of the most important tasks in modern biology. Genes rarely work in isolation. Instead, they operate within large regulatory networks, where certain genes act like “switches” that turn others on or off depending on the needs of the cell. Understanding these networks is essential for uncovering how healthy cells function and what goes wrong in diseases such as cancer, neurological disorders, or immune system malfunctions.

A team of scientists from Shanghai Jiao Tong University has developed a new method called KEGNI (Knowledge graph-Enhanced Gene regulatory Network Inference). KEGNI is designed to reveal the hidden relationships between genes by using advanced computational tools and big data.

The key innovation behind KEGNI is how it combines RNA sequencing data with a knowledge graph. RNA sequencing provides a snapshot of which genes are active inside a single cell, while a knowledge graph organizes biological information into a network of related concepts, like a map that connects genes, proteins, and cellular processes. By using a type of artificial intelligence called a graph autoencoder, KEGNI can integrate these two sources of information to build a more accurate picture of how genes regulate each other.

Schematic overview of the KEGNI framework

KEGNI is a comprehensive framework that integrates a Masked Autoencoder (MAE) model and a Knowledge Graph Embedding (KGE) model. a A base GRN is constructed from scRNA-seq data of a specific cell population using k-nearest neighbors (k-NN). The MAE takes this base GRN as input and focuses on the reconstruction of masked node features. b A cell type-specific knowledge graph is built using KEGG pathway information and relevant cell type markers. The KGE model employs a contrastive learning approach with negative sampling to achieve knowledge graph embedding. Nodes in the graph represent genes. Black hollow circles indicate genes from single-cell profiles; gray solid circles represent genes from the knowledge graph; red solid circles denote negative samples. Common genes are those shared between the single-cell expression profiles and the cell type-specific knowledge graph

When the researchers tested KEGNI, they found that it outperformed other popular methods. It worked well not only with single-cell RNA sequencing data, but also when combined with scATAC-seq data, which captures information about how DNA is packaged and made accessible in the cell. This makes KEGNI particularly powerful because it can take advantage of multiple data types to strengthen predictions.

One of KEGNI’s most exciting capabilities is its ability to identify driver genes. These are genes that play central roles in controlling cell behavior and may act as “master switches” in critical biological processes. Pinpointing these genes can help researchers better understand disease mechanisms and highlight potential therapeutic targets.

Another advantage of KEGNI is its modular design, meaning that it can be adapted to different research contexts. Scientists can plug in various knowledge graphs depending on the biological system they want to study, whether that’s stem cell development, cancer progression, or immune system regulation.

By bringing together artificial intelligence, RNA sequencing, and curated biological knowledge, KEGNI offers researchers a powerful new way to investigate how genes interact and control cellular functions. As tools like this continue to improve, they will accelerate discoveries about the inner workings of life and may open new doors to precision medicine.