# A copula-infused graph neural network for cell type classification in single-cell RNA sequencing data

## TL;DR

This paper introduces a new method for classifying cell types in single-cell RNA sequencing data using a combination of copula theory and graph neural networks.

## Contribution

The novelty lies in integrating copula theory with graph neural networks to better capture gene dependencies and cell relationships.

## Key findings

- scCopulaGNN performs well on high-dimensional and noisy scRNA-seq data.
- The model outperforms existing methods in cell type classification tasks.
- It provides detailed insights into cellular diversity and function.

## Abstract

Cell-type classification from single-cell RNA sequencing (scRNA-seq) data is among the most important steps in understanding cellular heterogeneity and biological mechanisms. High dimensionality, sparsity, and noise in scRNA-seq data lead to significant computational and statistical challenges. To this end, we devise a copula-infused graph neural network for single cell type classification (scCopulaGNN). Our model marries the flexibility of copula theory with the strong representation-learning capabilities of graph neural networks. The copula framework naturally captures complex dependencies among genes and the GNN models structural relationships among cells. scCopulaGNN is evaluated on real and simulated datasets and we demonstrate it can handle high-dimensional data with well performance. The model is also compared with existing methods to illustrate the model’s ability to classification task. These results highlight scCopulaGNN potential as an effective tool for cell type classification in single-cell transcriptomics, providing more elaborate details about cellular diversity and function.