Understanding how different cells behave within the same tissue is a major goal in modern biology. Single cell RNA sequencing allows researchers to measure gene activity in individual cells, revealing important differences between cell types. Researchers at the Guilin University of Electronic Technology have developed a new computational method to better analyze this complex data.

One of the key challenges in RNA sequencing data analysis is clustering, which groups cells with similar gene expression patterns. Traditional clustering methods can miss subtle patterns, especially when dealing with large and complex datasets. Other approaches, such as biclustering, attempt to group both cells and genes at the same time, but often struggle with accuracy and scalability.

To overcome these limitations, the researchers introduced a new algorithm called scDBic. This method uses deep learning, specifically a type of model known as a deep autoencoder, to capture important features in gene expression data. By reducing noise and focusing on the most meaningful signals, the algorithm improves how cells are grouped.

The workflow of cell clustering phase

Step a, b, c, and d respectively represent preprocessing module, deep autoencoder module, constructing the SNN graph module, and Walktrap clustering for cells module.

The approach also includes a strategy to identify key genes that define each cell group. This is important because it not only shows how cells are related, but also highlights the biological drivers behind those differences. As a result, researchers can gain deeper insight into cellular behavior and function.

When tested on single cell RNA sequencing datasets, scDBic outperformed traditional clustering and biclustering methods. It was able to more accurately identify cell populations and the genes associated with them, even in high dimensional data.

Overall, this work demonstrates how combining RNA sequencing with advanced computational tools can improve our understanding of cellular diversity. Better clustering methods like scDBic will help researchers uncover new cell types, understand disease mechanisms, and identify potential therapeutic targets.

Availability – The source code and test data are freely available at GitHub (https://github.com/Xiaoqi-Tang/scDBic)