scLLM-DSC: LLM-Knowledge Enhanced Cross-Modal Deep Structural Clustering for Single-Cell RNA Sequencing

The article introduces scLLM-DSC, a novel framework for single-cell RNA sequencing (scRNA-seq) that integrates Large Language Models (LLMs) with deep structural clustering. This framework combines a Knowledge-Driven Semantic View from NCBI gene priors and Cell2Sentence embeddings with a Structure-Aware Topological View using a graph-guided encoder, enhanced by a cross-modal contrastive alignment mechanism. Benchmark results indicate that scLLM-DSC surpasses eleven existing methods in clustering accuracy, highlighting its potential for improving cell population identification and tissue heterogeneity resolution in biological research.