GOWDL: gene ontology-driven wide and deep learning model for cell typing of scRNA-seq data

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) provides high-resolution cellular data.
• Accurate cell type identification is crucial for understanding tissue composition.
• Existing methods struggle with the scale and complexity of scRNA-seq data.

Purpose of the Study:

• To develop an automated deep learning model for cell type classification.
• To leverage Gene Ontology functional annotations and marker genes for improved accuracy.
• To evaluate the model's performance against state-of-the-art methods.

Main Methods:

• Implementation of a hybrid wide and deep learning architecture (GOWDL).
• Integration of Gene Ontology (GO) functional data and cell-type-specific marker genes.
• Cross-validation and independent external testing across five tissue types.

Main Results:

• GOWDL achieved superior classification performance across multiple tissues.
• The model demonstrated high accuracy, outperforming 12 other predictors.
• Recall was high (92%), slightly below the top-performing tool (97%).

Conclusions:

• GOWDL offers a powerful and accurate approach for cell type classification from scRNA-seq data.
• The model's hybrid architecture effectively combines functional and marker gene information.
• Demonstrated utility in a case study of immune cell classification in breast cancer.