Fereshteh Fallah Atanaki Leila Mirsadeghi Kaveh Kavousi ^*

• Laboratory of Complex Biological Systems and Bioinformatics, Institute of Biochemistry and Biophysics, University of Tehran, Tehran, Alborz, Iran

Tumor microenvironments (TMEs) encompass intricate ecosystems comprising cancer cells, infiltrating immune cells, and diverse cell types. Complex networks of intercellular and intracellular signals within the TME dictate cancer progression and therapeutic response. Various computational tools exist to study these interactions. However, incorporating tumor progression explicitly within these models across different cancer types remains a challenge. This study introduces a comprehensive framework that utilizes single-cell RNA sequencing (scRNA-seq) data within a multilayer network model. This enables a thorough investigation of molecular changes across glioma progression stages. The proposed heterogeneous, multilayered network model carefully replicates the biological system's hierarchical structure, from genetic building blocks through cellular functions to phenotypic manifestations. This model is designed to reveal previously hidden relationships among the microscopic constituents of the biological landscape, offering a deeper understanding into the cellular machinery. The proposed pipeline was applied to glioma scRNA-seq data, and complex network analysis of the distinct networks from different cancer stages revealed significant ligand‒receptor connections and the most crucial ligand‒receptor-transcription factor (TF) axes, along with their associated biological pathways. Differential network analysis conducted between grade III glioma and grade IV glioma uncovered the most important nodes and edges involved in rewiring interactions. Additionally, biological pathway enrichment analysis identified four genes containing one ligand (PDGFA), one receptor (PDGFRA), one TF (CREB1), and one target gene (PLAT) are involved in Signaling by Receptor Tyrosine Kinases (RTK) signaling pathways, which plays a pivotal role in the progression of grade III to grade IV glioma. These genes emerged as crucial features for machine learning in stage prediction, achieving 87% accuracy and 93% AUC in predicting 3-year survival via Kaplan-Meier analysis.