scATAC-seq preprocessing and imputation evaluation system for visualization, clustering and digital footprinting.

Area of Science:

• Genomics
• Epigenetics
• Computational Biology

Background:

• Single-cell ATAC-seq (scATAC-seq) enables chromatin accessibility and transcription factor binding analysis at a single-cell resolution.
• Data sparsity in scATAC-seq necessitates imputation, while large feature sets require preprocessing for computational efficiency.
• Optimal combined strategies for scATAC-seq imputation and preprocessing remain underexplored.

Purpose of the Study:

• To benchmark imputation frameworks for scATAC-seq data.
• To evaluate the combined effects of imputation and preprocessing on downstream analyses.
• To identify optimal preprocessing-imputation strategies for various scATAC-seq applications.

Main Methods:

• Development of SAPIEnS (scATAC-seq Preprocessing and Imputation Evaluation System).
• Integration of state-of-the-art imputation methods with common preprocessing techniques.
• Assessment of clustering, visualization, and digital genomic footprinting using benchmarked strategies.

Main Results:

• The Boruta preprocessing method proved beneficial across most scATAC-seq analysis tasks.
• Imputation strategies were found to be most effective for smaller datasets.
• Optimal strategies were identified based on analysis type and dataset feature count.

Conclusions:

• SAPIEnS provides a framework for evaluating scATAC-seq data processing pipelines.
• Preprocessing is generally more impactful than imputation, especially for larger datasets.
• The study offers guidance on selecting imputation and preprocessing methods for specific scATAC-seq research questions.