DrImpute

DrImpute imputes dropout events in single-cell RNA sequencing (scRNA-seq) data to distinguish dropout zeros from true zeros and improve downstream analyses.

Key Features:

• Dropout imputation: Imputes dropout events in scRNA-seq to recover missing expression values.
• Zero classification: Distinguishes dropout zeros from true biological zeros.
• Implementation: Implemented in R.
• Downstream improvement: Enhances accuracy of cell type identification, visualization, lineage reconstruction, and clustering.
• Compatibility with statistical tools: Improves performance of PCAreduce, SC3, PCA, t-SNE, Monocle, and TSCAN.
• Validation: Demonstrated across nine published scRNA-seq datasets.
• Performance: Demonstrated to outperform existing algorithms in separating dropout zeros from true zeros.

Scientific Applications:

• Cell type identification: Improves clustering and classification of single cells based on corrected expression profiles.
• Visualization: Enhances low-dimensional visualizations such as PCA and t-SNE by reducing technical zeros.
• Lineage and trajectory reconstruction: Improves reconstruction of lineage trajectories using tools like Monocle and TSCAN.
• Clustering and dimensionality reduction: Refines results from methods such as PCAreduce and SC3 by supplying imputed data.

Methodology:

Imputation of dropout events and classification of dropout versus true zeros in scRNA-seq data, implemented in R.