Are dropout imputation methods for scRNA-seq effective for scATAC-seq data?

Area of Science:

• Genomics
• Computational Biology
• Epigenetics

Background:

• Single-cell sequencing technologies enable high-resolution studies of cellular processes.
• Single-cell ATAC-seq (scATAC-seq) analyzes genome-wide chromatin accessibility.
• Dropout events in scATAC-seq data complicate downstream analyses like clustering.

Purpose of the Study:

• To evaluate the effectiveness of existing scRNA-seq imputation methods for inferring dropout events in scATAC-seq data.
• To systematically benchmark selected imputation methods using various downstream analyses.

Main Methods:

• Selected state-of-the-art scRNA-seq imputation methods: MAGIC, SAVER, scImpute, deepImpute, PRIME, bayNorm, and knn-smoothing.
• Applied these methods to infer dropout peaks in scATAC-seq data.
• Benchmarked performance based on correlation with meta-cell, clustering, subpopulation distance, imputation accuracy, TF motif identification, and computation time.

Main Results:

• Most imputed peaks enhanced correlation with the reference meta-cell.
• Method performance varied significantly across different datasets and downstream analyses.
• MAGIC demonstrated the most consistent superior performance across all evaluated metrics.

Conclusions:

• Imputation methods can improve scATAC-seq data quality, but their application requires caution due to dataset-specific performance variations.
• MAGIC is a reliable choice for inferring dropout events in scATAC-seq data.
• Source code is available for reproducible research.