The integration of multi-omics data across multiple cancer types, including Breast Invasive Carcinoma (BRCA), Lung Adenocarcinoma (LUAD) and Cervical Squamous Cell Carcinoma and Endocervical Adenocarcinoma (CESC) enriches with a better understanding of cancer biology. Three different types of data (mRNA, miRNA, and RPPA) for each cancer were integrated with the DIABLO and NOLAS methods. The study aimed to evaluate and compare the performance of both methods and study their strengths and limitations. The DIABLO model showed high specificity across datasets, having strong discrimination ability, though it had low sensitivity, which indicated difficulties of the model in the identification of tumor cases. Conversely, the NOLAS model showed a high robustness in handling data with noise and batch effects, which implied that the NOLAS model had better abilities for class separation. Both models had maintained moderate accuracy while making predictions for test data, and were effective in identifying potential biomarkers and relevant pathways. The DIABLO model selected a limited number of highly informative features, while the NOLAS model extracted a wide range of biological variations. Pathway analysis revealed the engagement of pathways related to the identified biomarkers by both models in cancer progression, such as "microRNAs in cancer" and "hedgehog signaling" for lung cancer and "epidermis development" for cervical cancer. The study showed both strengths and limitations of both models and suggested the need to improve the feature selection process and to solve issues with class imbalance to improve the prediction performance of both models. The research focused on improving multi-omics integration methods to contribute to cancer research and develop strategies for personalized medicine.