SigPrimedNet: A Signaling-Informed Neural Network for scRNA-seq Annotation of Known and Unknown Cell Types.

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) reveals cellular complexity.
• Manual cell annotation is infeasible due to data volume and heterogeneity.
• Automated methods are crucial for scRNA-seq data analysis.

Purpose of the Study:

• To develop an artificial neural network for efficient and accurate cell type annotation.
• To enable identification of both known and unknown cell types in scRNA-seq data.
• To provide insights into cell functionality through learned representations.

Main Methods:

• Introduced SigPrimedNet, an artificial neural network.
• Utilized a sparsity-inducing signaling circuits-informed layer for efficient training.
• Employed supervised training for feature representation learning.
• Integrated anomaly detection for unknown cell-type identification.

Main Results:

• SigPrimedNet accurately annotates known cell types with a low false-positive rate for unseen cells.
• Demonstrated effectiveness across multiple public scRNA-seq datasets.
• Learned representations serve as proxies for signaling circuit activity.

Conclusions:

• SigPrimedNet offers an efficient solution for automated cell type annotation in scRNA-seq data.
• The method successfully distinguishes known and novel cell populations.
• The learned representations enhance understanding of cell functionality and communication.