BiAEImpute: a robust bidirectional autoencoder framework for High-fidelity dropout imputation in single-cell...

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) provides deep insights into cellular heterogeneity.
• Dropout events, where transcripts go undetected, are a major challenge in scRNA-seq analysis.
• These missing values compromise the accuracy of downstream analyses.

Purpose of the Study:

• To develop an effective imputation method for scRNA-seq data.
• To mitigate the adverse effects of dropout events on data analysis.

Main Methods:

• A bidirectional autoencoder-based model, BiAEImpute, was developed.
• The model utilizes row-wise and column-wise autoencoders to learn cellular and genetic features.
• Synergistic integration of learned features enables robust and accurate imputation.

Main Results:

• BiAEImpute demonstrated superior performance on four real scRNA-seq datasets compared to existing methods.
• The model successfully restored missing values.
• BiAEImpute improved cell subpopulation clustering, marker gene identification, and developmental trajectory inference.

Conclusions:

• BiAEImpute is an effective and resilient tool for imputing missing data in scRNA-seq.
• The method enhances the accuracy of downstream scRNA-seq analyses.
• Source code is publicly available for use and further development.