Computational methods for therapeutic target discovery face challenges in integrating multi-scale biological data and predicting system-wide therapeutic effects. We present a computational framework that integrates single-cell transcriptomics, weighted gene co-expression network analysis (WGCNA), and computational perturbation simulation to systematically discover novel therapeutic targets. The framework constructs a knowledge graph comprising 29 896 nodes (23 530 genes and 6366 pathways) with 322 136 edges, integrating gene-gene, gene-pathway, and module relationships. Using perturbation simulation algorithms, we systematically explored 265 candidate targets, scoring each based on perturbation magnitude, module response, therapeutic effect, and statistical significance. Applied to single-cell RNA sequencing data from 38 patients (51 415 cells), the framework identified 66 novel therapeutic targets, including nine very high novelty targets. Computational validation demonstrates efficient scalability (knowledge graph construction: <5 min; 265 perturbation simulations: <2 min) and robust performance across different module sizes. This approach represents a novel computational method for systems-level therapeutic target discovery, with generalizable applications to other complex diseases.
Sun et al. (Sun,) studied this question.