MarkerCapsule

MarkerCapsule applies capsule networks to single-cell RNA sequencing (scRNA-seq) data to automate and interpret cell-type identification.

Key Features:

• Automated Annotation: Automates cell-type annotation from scRNA-seq data to reduce manual curation and bias.
• Integration of Heterogeneous Data: Integrates diverse datasets for coherent single-cell analyses.
• Biologically Meaningful Interpretation: Associates marker genes with capsule network units to provide interpretable gene–cell-type relationships.
• Efficiency with Limited Labeled Data: Operates effectively with a small amount of labeled data.
• Superior Typing Accuracy: Demonstrates higher cell-typing accuracy compared to existing state-of-the-art methods.

Scientific Applications:

• Genomics and Single-cell Biology: Supports cell-type identification and characterization in genomics and single-cell biology studies.
• Cellular Heterogeneity Analysis: Enables exploration of cellular heterogeneity within complex tissues.
• Novel Cell-type Discovery: Facilitates identification of novel cell types and subtypes from scRNA-seq data.
• Gene Expression Pattern Characterization: Provides insights into marker gene activity patterns that define cell populations.

Methodology:

MarkerCapsule employs capsule networks from deep learning to model relationships between marker genes and cell types by capturing spatial hierarchies among features.