A novel multi-objective optimization framework using NSGA-II for gene co-expression network inference

Gene co-expression networks (GCNs) provide a powerful framework for uncovering functional gene modules and biological pathways from complex transcriptomic data. However, constructing reliable GCNs from noisy datasets often yields spurious edges and biologically implausible topologies. To address this challenge, we propose a novel multi-objective optimization approach based on the Non-dominated Sorting Genetic Algorithm II (NSGA-II) to refine edge selection in GCNs. Our pipeline integrates Variance Stabilizing Transformation (VST) for RNA-seq normalization, Spearman rank correlation for robust co-expression estimation, permutation testing to establish an initial significance threshold, and bootstrap resampling to assess edge stability. We applied this framework to two heterogeneous datasets including GSE10245 (microarray, n = 58) and GSE102349 (RNA-seq, n = 113), to optimize multiple network properties simultaneously; including sparsity, modularity, scale-free topology, and edge reproducibility. Comparative analyses against conventional widely used methods; Weighted Gene Co-expression Network Analysis (WGCNA) and the Algorithm for the Reconstruction of Accurate Cellular Networks (ARACNE), demonstrate that our approach consistently yields sparser, more modular networks that better conform to biologically expected scale-free architectures across both data types. This adaptive, optimization-driven strategy offers a robust foundation for integrative genomic studies and holds significant potential for advancing biomarker discovery and disease mechanism modeling.