ComicGTN infers disease-associated rare cell states from single-cell multiomic data using DNA sequence-augmented...

Area of Science:

• Genomics and Computational Biology
• Single-cell multiomics analysis
• Epigenetics and Transcriptomics

Background:

• Single-cell multiomic technologies enable simultaneous profiling of transcriptome and epigenome.
• Identifying rare cell states is critical for discovering disease biomarkers.
• Existing methods overlook DNA sequence information at accessible sites.

Purpose of the Study:

• To develop an innovative computational framework, ComicGTN, for accurate identification of rare cell clusters.
• To integrate single-cell multiomic data with DNA sequence information.
• To improve the detection of disease-associated rare cell populations.

Main Methods:

• Developed ComicGTN, a computational framework utilizing enhanced graph transformer networks.
• Integrated single-cell transcriptome, epigenome, and DNA sequence data.
• Evaluated performance against nine state-of-the-art methods across multiple datasets.

Main Results:

• ComicGTN significantly outperformed existing methods in identifying rare cells.
• Detected distinct immune subpopulations in mouse breast cancer.
• Uncovered specific oligodendrocyte differentiation states in epilepsy patient cortex.
• Elaborated pathological pathways in a rare glomerular subgroup from polycystic kidney disease organoids.

Conclusions:

• ComicGTN effectively identifies rare cell populations using integrated multiomic and sequence data.
• The framework accelerates the localization of disease-associated cells.
• ComicGTN facilitates clinical insights into development and disease progression.