ADImpute

ADImpute predicts missing expression values in single-cell RNA sequencing data through adaptive dropout imputation that integrates novel gene regulatory network-based methods with existing imputation approaches.

Key Features:

• Gene Regulatory Network-Based Imputation: Uses gene-gene relationships learnt from external data to predict missing values.
• Baseline Average Method: Implements a sample-wide average approach as a baseline imputation method.
• Integration of Existing Methods: Combines novel methods with existing imputation methods including DrImpute and SAVER.
• Adaptive Ensemble Learning: Learns the best performing method per gene and combines results from different methods into an ensemble.

Scientific Applications:

• Single-Cell RNA Sequencing Analysis: Addresses dropout events where scRNA-seq methods cannot quantify expression levels of all genes in a cell.
• Gene Expression Recovery: Exploits external gene-gene relationship information to improve imputation beyond dataset-only approaches.

Methodology:

ADImpute implements a gene regulatory network-based approach using external gene-gene relationships, a sample-wide average baseline approach, and combines these with existing methods (DrImpute, SAVER) through per-gene adaptive learning to generate ensemble predictions.