A novel f -divergence based generative adversarial imputation method for scRNA-seq data analysis.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) is crucial for understanding cellular diversity and developing personalized therapies.
• Missing values, or dropouts, in scRNA-seq data present significant analytical challenges.
• Traditional imputation methods often rely on restrictive distributional assumptions and perform poorly at high missing rates.

Approach:

• We introduce sc-f GAIN, a novel generative adversarial imputation method utilizing f-divergence for scRNA-seq data.
• sc-f GAIN integrates four f-divergence functions (cross-entropy, KL, reverse KL, Jensen-Shannon) into a generative adversarial network.
• This approach generates imputed values without distributional assumptions and preserves the original data distribution.

Key Points:

• sc-f GAIN demonstrates superior imputation performance compared to traditional methods, evidenced by lower root-mean-square error.
• The method exhibits robustness across varying missing rates and effectively reduces imputation bias.
• The f-divergence framework offers flexibility, enabling sc-f GAIN to handle diverse data types.

Conclusions:

• sc-f GAIN provides a universal and robust solution for imputing missing values in scRNA-seq data.
• Accurate imputation enhances the analysis of cellular heterogeneity and supports precision medicine initiatives.
• This advanced imputation technique holds promise for broader applications in transcriptomic data analysis.