iCellR

iCellR performs batch alignment and analysis of single-cell sequencing data, including scRNA-seq, scVDJ-seq, and CITE-seq, to mitigate drop-outs from under-sampling and low-coverage sequencing and improve dimensionality reduction, clustering, and downstream analyses.

Key Features:

• Combined Coverage Correction Alignment (CCCA): Employs a coverage correction approach akin to imputation to harmonize gene expression matrices across multiple samples and correct drop-outs.
• Combined Principal Component Alignment (CPCA): Aligns principal components across samples using k nearest neighbors (KNN) without performing coverage correction.
• Drop-out mitigation: Targets zero counts arising from under-sampling RNA molecules and low-coverage sequencing to reduce masking of true biological signal.
• Cell cycle analysis: Provides cell cycle analysis and scoring across phases G0, G1S, G2M, M, G1M, and S.
• Pseudotime Abstract KNetL map (PAK map): Includes a PAK map implementation for pseudotime analysis.
• Gene-gene correlations: Computes gene-gene correlation metrics for single-cell expression data.
• PBMC demonstration: Methods have been demonstrated on nine scRNA-seq peripheral blood mononuclear cell (PBMC) samples from various batches and technologies.

Scientific Applications:

• Multi-modal single-cell integration: Integration and joint analysis of scRNA-seq, scVDJ-seq, and CITE-seq datasets across batches and technologies.
• Batch correction and clustering: Reduction of batch effects and improvement of dimensionality reduction and clustering in single-cell studies.
• Cell cycle assignment: Identification and analysis of cell cycle phases (G0, G1S, G2M, M, G1M, S) in single cells.
• Pseudotime ordering: Pseudotime inference and trajectory analysis using the PAK map.
• Gene interaction analysis: Exploration of gene-gene correlation structures within single-cell datasets.
• Benchmarking on PBMCs: Evaluation of batch-alignment performance on PBMC scRNA-seq samples spanning multiple batches and technologies.

Methodology:

Implemented in R; CCCA performs joint coverage correction analogous to imputation to harmonize expression matrices and correct drop-outs; CPCA aligns principal components across samples via KNN without coverage correction; includes cell cycle phase scoring, PAK map computation, and gene-gene correlation calculations.