PACells identifies phenotype-associated cell states from single-cell chromatin accessibility profiles

Area of Science:

• Genomics
• Computational Biology
• Epigenetics

Background:

• Identifying cell states driving disease is key for understanding pathogenesis and developing targeted therapies.
• Single-cell assay for transposase-accessible chromatin using sequencing (scATAC-seq) provides high-resolution insights into cellular regulatory landscapes.
• Linking bulk ATAC-seq clinical data with scATAC-seq cellular data remains a challenge.

Purpose of the Study:

• To present PACells, a novel computational framework for integrating bulk and single-cell ATAC-seq data.
• To identify critical cell states associated with clinical phenotypes at single-cell resolution.
• To enhance the understanding of disease mechanisms and identify potential therapeutic targets.

Main Methods:

• PACells framework development for linking bulk ATAC-seq clinical data with scATAC-seq data.
• Benchmarking PACells against existing methods for cell prediction and molecular signature identification.
• Application of PACells to Alzheimer disease and glioblastoma datasets.
• Extension of PACells to transcriptomics data for melanoma immunotherapy outcomes.

Main Results:

• PACells demonstrates superior performance in predicting cells and molecular signatures associated with disease and gene mutations compared to other methods.
• The framework successfully identified clinical cell states and regulatory elements relevant to Alzheimer disease.
• PACells discerned cell states linked to poor survival in glioblastoma.
• The extended PACells framework showed utility in analyzing melanoma transcriptomics data for immunotherapy outcomes.

Conclusions:

• PACells provides a powerful tool for dissecting cellular heterogeneity and its link to clinical phenotypes using ATAC-seq data.
• The framework advances the identification of disease-specific cell states and regulatory elements, paving the way for precision medicine.
• PACells has broad applicability across different diseases and data modalities, including transcriptomics.