AI-Driven Multi-Omics Integration for Targeted Drug Discovery Against Cancer Stem Cells: A Genomic Perspective

Abstract

Cancer stem cells (CSCs) are responsible for the initiation of tumors, their relapses and drug resistance to a very high extent, and the molecular heterogeneity between these cells is rarely identified by single-omics analysis. Artificial intelligence (AI) can be very useful for the integration of genomic, transcriptomic, epigenomic, and proteomic data, but it is less known if scientists will value this research. This study examines how much researchers perceive the effectiveness of AI-driven multi-omics integration for the clinical applicability in CSC-directed drug discovery. A structured questionnaire of 15 items was administered to 63 respondents from [bioinformatics, biotechnology, oncology and allied fields], and the data were analyzed using partial least squares structural equation modelling (PLS-SEM) in SmartPLS 4. All constructs showed acceptable internal consistency (Cronbach's α = 0.785–0.813; composite reliability = 0.853–0.869; AVE = 0.542–0.576) and no collinearity concerns (VIF ≤ 3.307). AI-driven multi-omics integration was strongly associated with perceived classification capability (β = 0.833, R² = 0.694) and with explainability and clinical applicability (β = 0.875, R² = 0.765). However, since heterotrait–monotrait ratios are above 1.0, it reveals that respondents did not clearly separate the three constructs, suggesting they perceive AI integration, classification and explainability as a single, unified value proposition. The findings indicate that researchers have strong confidence in integrated AI approaches for precision oncology, while underlining the need for sharper construct measurement and empirical benchmarking of AI models on real multi-omics data.