The rapid growth of omics and ontology data, coupled with advancements in large language models (LLMs) and deep learning, has opened new opportunities for data-driven drug target and biomarker discovery. By integrating these technologies, we aim to streamline cancer research through comprehensive data mining and network-based analysis.

We integrated diverse omics and ontology datasets from public sources into the Q-omics data mining platform (https://qomics.io). Meta-analysis and data consensus were performed across datasets and incorporated into Q-omics' smart databases. A LLM-powered "Text-to-Data Mining" interface was implemented, enabling researchers to perform complex analyses without computational expertise. For functional insights, we developed NetCrafter (https://netcrafter.sookmyung.ac.kr), a novel ontology-based network analysis algorithm, and incorporated it into the Q-omics workflow. Graph neural networks (GNNs) such as message-passing neural networks (MPNNs) were used to identify network hotspots, facilitating drug target discovery.

We identified over 29 billion significant cross-associations across heterogeneous datasets, encompassing over 1 billion multi-modal omics data points. Consensus analysis on cross-associated data across pan-cancer datasets provides a valuable resource for identifying reliable drug targets, biomarkers, and underlying mechanisms. Genes with tumor-specific overexpression and unfavorable prognoses show high consistency across cancer lineages. Lineage consensus in gene perturbation data (CRISPR/shRNA) demonstrates strong reproducibility, reinforcing the reliability of functional screening approaches. In addition, advanced data mining applications within Q-omics enable the identification of synthetic lethal gene pairs and tumor-specific neoantigens, supporting novel therapeutic strategies. Ontology-based network analysis provides immediate functional insights from gene lists derived from RNA/protein expression or perturbation experiments (CRISPR/shRNA). Ontological networks highlight pan-cancer consensus functions associated with drug response and patient survival and serve as valuable input for MPNN models, aiding hotspot identification for drug target discovery.

By integrating multi-omics meta-data with LLMs and deep learning, Q-omics pioneers an innovative "Text-to-Data Mining" platform, enabling researchers to effortlessly identify anticancer targets, biomarkers, and mechanistic insights—without computational expertise. As omics datasets expand and AI technology advances, Q-omics (https://qomics.io) serves as a transformative resource, driving progress in cancer research and drug development.