devCellPy is a machine learning-enabled pipeline for automated annotation of complex multilayered single-cell...

Area of Science:

• Computational Biology
• Developmental Biology
• Genomics

Background:

• Single-cell RNA sequencing (scRNA-seq) analysis faces informatic challenges in annotating cells with complex or transitional identities.
• Precise cell type annotation is crucial for understanding biological systems, especially during development.

Purpose of the Study:

• To introduce devCellPy, a machine learning tool for automated cell type prediction in scRNA-seq data.
• To demonstrate devCellPy's capability in handling complex annotation hierarchies and cross-species predictions.

Main Methods:

• Construction of a murine cardiac developmental atlas from published scRNA-seq datasets (104,199 cells, E6.5-E16.5).
• Training devCellPy with the murine cardiac atlas to generate a prediction algorithm.
• Cross-species prediction using human induced pluripotent stem cells (hiPSCs) and validation with lineage tracing.

Main Results:

• devCellPy achieved high prediction accuracy (>90%) across multiple annotation layers in murine cardiac development data.
• The tool demonstrated robust performance on de novo murine developmental datasets.
• Cross-species analysis revealed an unexpected predominance of left ventricular (LV) identity in hiPSC-derived cardiomyocytes, confirmed by TBX5 lineage tracing.

Conclusions:

• devCellPy is a highly accurate and precise tool for automated cell type prediction in scRNA-seq data.
• The tool effectively handles complex cellular hierarchies, developmental atlases, and cross-species applications.
• devCellPy facilitates biological discovery by enabling robust cell annotation across diverse experimental systems.