Deep autoencoder for interpretable tissue-adaptive deconvolution and cell-type-specific gene analysis.

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) offers high-resolution biological insights but faces technical limitations with emerging data.
• Accurate deconvolution of bulk RNA sequencing (bulk RNA-seq) data is crucial for understanding tissue composition and cellular heterogeneity.

Purpose of the Study:

• To introduce Tissue-AdaPtive autoEncoder (TAPE), a deep learning method for precise and rapid deconvolution of bulk RNA-seq data.
• To enable tissue-adaptive prediction of cell-type fractions and cell-type-specific gene expression profiles.

Main Methods:

• TAPE utilizes a deep learning framework to integrate bulk RNA-seq and scRNA-seq data.
• An interpretable decoder and a unique training scheme are employed for accurate predictions.
• The method is validated on multiple datasets and compared against existing popular deconvolution techniques.

Main Results:

• TAPE demonstrates superior overall performance and comparable cell-type level accuracy compared to existing methods.
• The method exhibits robustness across different cell types, enhanced speed, and sensitivity in generating biologically meaningful predictions.
• Analysis of clinical data confirms TAPE's capability in predicting biologically significant cell-type-specific gene expression profiles.

Conclusions:

• TAPE provides a powerful and efficient tool for precise deconvolution of RNA sequencing data.
• The method facilitates accelerated and accurate analysis of high-throughput clinical data, advancing biological discovery.
• TAPE enhances the application of scRNA-seq principles to bulk RNA-seq data, offering significant biological insights.