Skip to main content

Single-cell Graph Convolutional Network allows for rapid knowledge transfer

Jing Su
An IU School of Medicine researcher developed a graph artificial intelligence model for single cell omics data analysis. Jing Su, PhD, is one of the corresponding authors of a Nature Communications publication about the single-cell Graph Convolutional Network (scGCN), which can help researchers transfer knowledge across different studies and datasets. scGCN software is currently available for researchers to download and use.  

“Single-cell omics is the fastest-growing type of genomics data in the biomedical research and public genomics repositories,” said Su, assistant professor of biostatistics and health data science at IU School of Medicine. “Leveraging the growing repository of single-cell omics datasets which has been well-annotated for cell types and biological activities and transferring such knowledge in terms of annotations from existing datasets to newly generated datasets will empower the exploration of single-cell omics data."  

Su was recruited to the Biostatistics and Health Data Science Department in 2020 as an IU Precision Health Initiative faculty member. Data and informatics comprise one of the scientific pillars of the program, which has been implemented cross-functionally to focus on diseases such as triple-negative breast cancer and Alzheimer’s disease. Su said their novel graph artificial intelligence tool can help understand the complexity and diversity of cells, as well as their interactions in cancers and Alzheimer’s disease.

“It can tell us a lot about how cancer cells leverage the environment and modulate the immune system toward their benefit,” said Su. “It also can help us understand how treatments and new therapies can undermine such cellular interactions and better treat cancers. That’s one unique use of this type of technology.”

By using graph artificial intelligence, Su said they can efficiently handle huge amounts of data – including information from about millions of different cells – through splitting them into smaller subgraphs, divide, and conquer. Thus, scGCN can process big data very quickly, rather than making researchers have to wait dozens of hours on supercomputers, as they have had to do in the past. Thanks to this unique property of graphs, scGCN can also process big data on a regular laptop instead of supercomputers. The software will also allow researchers to translate information from animal models to human cells, which will help them map the knowledge from preclinical studies to clinical research at the single cell level. This capability can ultimately help speed up drug development approaches in the future.

The views expressed in this content represent the perspective and opinions of the author and may or may not represent the position of Indiana University School of Medicine.

Anna Carrera

Research Communications Manager

Anna Carrera is the research communications manager for Indiana University's Precision Health Initiative, IU School of Medicine and the Indiana Clinical and Translational Sciences Institute. She joined the team in June 2019 after working as a TV news rep...