Multi-omics data analysis is an important aspect of cancer molecular biology studies and has led to ground-breaking discoveries. Many efforts have been made to develop machine learning methods that automatically integrate omics data. Here, we review machine learning tools categorized as either general-purpose or task-specific, covering both supervised and unsupervised learning for integrative analysis of multi-omics data. We benchmark the performance of five machine learning approaches using data from the Cancer Cell Line Encyclopedia, reporting accuracy on cancer type classification and mean absolute error on drug response prediction, and evaluating runtime efficiency. This review provides recommendations to researchers regarding suitable machine learning method selection for their specific applications. It should also promote the development of novel machine learning methodologies for data integration, which will be essential for drug discovery, clinical trial design, and personalized treatments.

Recurrent mutations in SPOP-encoding a Cullin 3-based E3 ubiquitin ligase- in prostate cancer disrupt the recognition and degradation of ubiquitination substrates, including BET proteins. Consequently, stability of BET proteins is enhanced and this increases the resistance to BET inhibitors in SPOP-mutant prostate tumors. These results, together with those in Janouskova et al. and Zhang et al., uncover a novel non genetic mechanism of resistance to BET inhibition involving cancer type-specific mutations in SPOP, and support the evaluation of SPOP mutations to inform the administration of BET inhibitors in the clinic. The bromodomain and extraterminal (BET) family of proteins comprises four members—BRD2, BRD3, BRD4 and the testis-specific isoform BRDT—that largely function as transcriptional coactivators1,2,3 and play critical roles in various cellular processes, including the cell cycle, apoptosis, migration and invasion4,5. BET proteins enhance the oncogenic functions of major cancer drivers by elevating the expression of these drivers, such as c-Myc in leukemia6,7, or by promoting the transcriptional activities of oncogenic factors, such as AR and ERG in prostate cancer8. Pathologically, BET proteins are frequently overexpressed and are clinically linked to various types of human cancer5,9,10; they are therefore being pursued as attractive therapeutic targets for selective inhibition in patients with cancer. To this end, a number of bromodomain inhibitors, including JQ1 and I-BET, have been developed11,12 and have shown promising outcomes in early clinical trials. Although resistance to BET inhibitors has been documented in preclinical models13,14,15, the molecular mechanisms underlying acquired resistance are largely unknown. Here we report that cullin-3SPOP earmarks BET proteins, including BRD2, BRD3 and BRD4, for ubiquitination-mediated degradation. Pathologically, prostate cancer–associated SPOP mutants fail to interact with and promote the degradation of BET proteins, leading to their elevated abundance in SPOP-mutant prostate cancer. As a result, prostate cancer cell lines and organoids derived from individuals harboring SPOP mutations are more resistant to BET-inhibitor-induced cell growth arrest and apoptosis. Therefore, our results elucidate the tumor-suppressor role of SPOP in prostate cancer in which it acts as a negative regulator of BET protein stability and also provide a molecular mechanism for resistance to BET inhibitors in individuals with prostate cancer bearing SPOP mutations.

The rapid advancement of high-throughput sequencing and other assay technologies has resulted in the generation of large and complex multi-omics datasets, offering unprecedented opportunities for advancing precision medicine. However, multi-omics data integration remains challenging due to the high-dimensionality, heterogeneity, and frequency of missing values across data types. Computational methods leveraging statistical and machine learning approaches have been developed to address these issues and uncover complex biological patterns, improving our understanding of disease mechanisms. Here, we comprehensively review state-of-the-art multi-omics integration methods with a focus on deep generative models, particularly variational autoencoders (VAEs) that have been widely used for data imputation, augmentation, and batch effect correction. We explore the technical aspects of VAE loss functions and regularisation techniques, including adversarial training, disentanglement, and contrastive learning. Moreover, we highlight recent advancements in foundation models and multimodal data integration, outlining future directions in precision medicine research.

The proteome provides unique insights into disease biology beyond the genome and transcriptome. A lack of large proteomic datasets has restricted the identification of new cancer biomarkers. Here, proteomes of 949 cancer cell lines across 28 tissue types are analyzed by mass spectrometry. Deploying a workflow to quantify 8,498 proteins, these data capture evidence of cell-type and post-transcriptional modifications. Integrating multi-omics, drug response, and CRISPR-Cas9 gene essentiality screens with a deep learning-based pipeline reveals thousands of protein biomarkers of cancer vulnerabilities that are not significant at the transcript level. The power of the proteome to predict drug response is very similar to that of the transcriptome. Further, random downsampling to only 1,500 proteins has limited impact on predictive power, consistent with protein networks being highly connected and co-regulated. This pan-cancer proteomic map (ProCan-DepMapSanger) is a comprehensive resource available at https://cellmodelpassports.sanger.ac.uk.

Formalin‐fixed, paraffin‐embedded (FFPE), biobanked tissue samples offer an invaluable resource for clinical and biomarker research. Here we developed a pressure cycling technology (PCT)‐SWATH mass spectrometry workflow to analyze FFPE tissue proteomes and applied it to the stratification of prostate cancer (PCa) and diffuse large B‐cell lymphoma (DLBCL) samples. We show that the proteome patterns of FFPE PCa tissue samples and their analogous fresh frozen (FF) counterparts have a high degree of similarity and we confirmed multiple proteins consistently regulated in PCa tissues in an independent sample cohort. We further demonstrate temporal stability of proteome patterns from FFPE samples that were stored between one to 15 years in a biobank and show a high degree of the proteome pattern similarity between two types of histological regions in small FFPE samples, i.e. punched tissue biopsies and thin tissue sections of micrometer thickness, despite the existence of a certain degree of biological variations. Applying the method to two independent DLBCL cohorts we identified myeloperoxidase (MPO), a peroxidase enzyme, as a novel prognostic marker. In summary, this study presents a robust proteomic method to analyze bulk and biopsy FFPE tissues and reports the first systematic comparison of proteome maps generated from FFPE and FF samples. Our data demonstrate the practicality and superiority of FFPE over FF samples for proteome in biomarker discovery. Promising biomarker candidates for PCa and DLBCL have been discovered.