Abstract— Lung adenocarcinoma (LUAD), a major subtype of non-small cell lung cancer, poses a persistent health challenge due to the lack of precise biomarkers that can guide early detection and treatment decisions. Major challenge in this multiomics integration lies in missing values and lack of paired patient samples, across different patient cohorts. In this study we combined miRNA and RNA expression profiles for LUAD patients. Applied GAIN to impute missing values without discarding the unpaired data. Our pathway level integration framework uses unpaired datasets, while maintaining its biological significance. LASSO based feature selection was implemented to identify stable genes associated with LUAD for RNA and miRNA separately. KEGG enrichment analysis of selected miRNA and RNA was performed and overlapping pathways were identified to select candidate genes. Five pathway supported candidate genes (CYP17A1, VANGL1, NRAS, PRICKLE2, and RAC1) demonstrated strong association with LUAD in LinkedOmics dataset, based on Kaplan Mier plot and cox proportional hazard models. External validation of these candidate genes was performed using two independent GEO datasets. PubMed literature count of the genes was also included to give biological relevance from literature. Overall, this framework gives robust biomarker discovery from unpaired data.