Deep learning-based approaches for integrating multi-omics data offer a novel perspective on cancer recurrence prediction. However, existing methods struggle to manage the complex relationships within multi-omics data and the intrinsic correlations between samples, leading to suboptimal prediction accuracy. To tackle these challenges, we propose a multi-omics cancer recurrence prediction model (MultCPM), which employs a multi-head attention mechanism to extract key information from biological pathways. Integrated with a hierarchical fusion module, the model performs layered integration of omics data to effectively capture their interdependence. Ultimately, the fused information is consolidated into a unified feature matrix, refining critical features and their relationships across omics. Results from 5-fold cross-validation, repeated five times on Breast Cancer (BRCA), Bladder Cancer (BLCA), and Liver Cancer (LIHC) datasets, demonstrate that the MultCPM model achieves superior prediction performance and robustness. Additionally, Deep SHapley Additive exPlanations (DeepSHAP) was utilized to analyze the model’s interpretability, revealing key genes closely associated with cancer recurrence, thus providing valuable insights for biological research and the development of cancer recurrence prediction algorithms. The code is publicly available at https://github.com/dowell2016/MultCPM .

Constraint-based modeling can mechanistically simulate the behavior of a biochemical system, permitting hypothesis generation, experimental design and interpretation of experimental data, with numerous applications, especially the modeling of metabolism. Given a generic model, several methods have been developed to extract a context-specific, genome-scale metabolic model by incorporating information used to identify metabolic processes and gene activities in each context. However, the existing model extraction algorithms are unable to ensure that a context-specific model is thermodynamically flux consistent. Here we introduce XomicsToModel, a semiautomated pipeline that integrates bibliomic, transcriptomic, proteomic and metabolomic data with a generic genome-scale metabolic reconstruction, or model, to extract a context-specific, genome-scale metabolic model that is stoichiometrically, thermodynamically and flux consistent. One of the key advantages of the XomicsToModel pipeline is its ability to seamlessly incorporate omics data into metabolic reconstructions, ensuring not only mechanistic accuracy but also physicochemical consistency. This functionality enables more accurate metabolic simulations and predictions across different biological contexts, enhancing its utility in diverse research fields, including systems biology, drug development and personalized medicine. The XomicsToModel pipeline is exemplified for extraction of a specific metabolic model from a generic metabolic model; it enables omics data integration and extraction of physicochemically consistent mechanistic models from any generic biochemical network. It can be implemented by anyone who has basic MATLAB programming skills and the fundamentals of constraint-based modeling.

The minor effects of many SNP interactions often determine complex traits. This interaction, known as epistasis, represents a non-additive genetic effect in which the influence of one variant depends on the presence of others. In this study, we tested for epistatic effects on the residual feed intake (RFI) and residual methane emission (RME) traits of Nelore cattle. Additionally, we evaluated the impact of these interactions in other omics layers (i.e., microorganism profiles in the rumen content and feces and mRNA and miRNA expression in the rumen wall). The genomic interaction modules identified 14 and 10 significant SNP-SNP modules associated with RME and RFI traits, respectively. The majority of these SNPs were located in intronic and intergenic regions. The top pathways and processes associated with the SNP-SNP modules were identified, with several pathways related to the immune system and actin cytoskeleton organization. Furthermore, many other omics data were correlated with these SNP-SNP modules. Our findings suggest that the immune response and cilium organization may play important roles in feed efficiency. These insights not only provide novel candidates for enhancing these traits through microbiota composition and transcriptional regulation but also underscore the power of network analysis in uncovering new functional interactions. This research provides new insights and highlights candidate features for improving cattle feed efficiency and methane emissions.

Toward the Best Generalizable Performance of Machine Learning in Modeling Omic and Clinical Data.

Our goal is to leverage publicly available whole transcriptome and genome-wide CRISPR-Cas9 screen data to identify and prioritize novel breast cancer therapeutic targets. We used DepMap dependency scores > 0.5 to identify genes that are potential therapeutic targets in 48 breast cancer cell lines. We removed genes that were pan-essential or were not expressed in TCGA breast cancer cohort. Genes were prioritized based on druggability using the Drug-Gene Interaction Database. Targets were defined separately for ER+, HER2+, and TNBC. A broader list of genes with dependency score > 0.25 were used to assess the associations between dependency scores and mutations and copy number variations (CNV) to identify potential synthetic lethal relationships and to map survival critical genes into biological pathways. 66, 53, and 29 genes were prioritized as targets in ER+, HER2+, and TNBC, respectively. These included known actionable targets and many novel targets. ER+ included FOXA1, GATA3, LDB1, TRPS1, NAMPT, WDR26, and ZNF217; HER2+ cancers included STX4, HECTD1, and TBL1XR1; and TNBC included GFPT1 and GPX4. Synthetic lethal associations revealed 5 and 19 significant associations between potential survival critical genes and mutations in HER2+ and TNBC, respectively. For example, PIK3CA mutation increased dependency on NDUFS3 in HER2+ cancers, and CNTRL mutation increased dependency on electron transport chain (ETC) genes in TNBC. 329, 747, and 622 CNVs showed synthetic lethal association in ER+, HER2+, and TNBC, respectively. We provide a genome-wide drug target prioritization list for breast cancer derived from integrated large-scale omics data.

Developing risk prediction models for type 2 diabetes and assessing the role of circulating metabolic biomarkers in five independent Finnish cohorts with over 22,000 individuals.

Patients with brain cancers are diagnosed based on MRI in the clinical setting while molecular signatures offer potential therapeutic targets. The necessity to re-form molecular and imaging information motivated our meta-analysis to decipher the correlation between the MRI-classified tumor locations, gene expression, and protein signatures in GBM. We analyzed spatial and omics data alongside the assessment of post-translational modifications. We first utilized MRI data to classify GBM into 4 groups. We then integrated imaging groups with RNA-Seq and proteomic data to determine the association between tumor locations, gene signatures, and protein abundance. Furthermore, we scrutinized independent measurements of post-translational modifications in each group of MRI-classified GBM. The coherent layer of imaging and molecular data collectively showed the dysregulation of cell cycle, ECM organization, immune infiltration or surveillance in all GBM cases regardless of tumor locations. Several neuronal and synaptic-specific genes were differentially altered, indicative of aberrant neuroactivity in GBM. These dysregulated genes and networks provided druggable targets that led to small compounds identification, possessing cytotoxicity against primary GBM and spanning histological boundaries. Our analysis also revealed lesion-specific molecular signatures in each group of GBM, suggesting pathological features uniquely in subgroups of GBM with prognostic or therapeutic potential. Moreover, alterations in post-translational modifications would be noteworthy to explore clinical applications. Deliverables from our meta-analysis hold the potential to inform therapeutic intervention. Despite heterogeneity in GBM, our findings implicate new directions of emerging treatments that may be used as concomitants to chemo-, radio- or immunological therapies.

Biomedical data visualization is critical for interpreting complex datasets, yet the clarity and quality of visualizations vary widely across tools and applications. This study introduces a comprehensive framework for evaluating biomedical figures and benchmarking visualization platforms. We developed Metrics for Evaluation and Discretization of Biomedical Visuals using an Iterative Scoring algorithm (M.E.D.V.I.S.), a quantification system that systematically assesses figure quality based on four criteria: Complexity, color usage, whitespace, and the number of distinct visualizations. The algorithm integrates dimensionality reduction, clustering, and thresholding to classify figures and generate tailored feedback for improvement. In parallel, we conducted a comparative analysis of 26 widely used visualization tools, evaluating each based on ease of use, customizability, financial cost, and required background knowledge. To demonstrate real-world applicability, we present case studies on figure evaluation in published research and introduce SpatioView, an interactive, browser-based platform for exploring spatial omics data. Collectively, our findings highlight the need for standardized evaluation methods and provide accessible solutions for improving figure design in biomedical research, education, and industry.

Biomedical data visualization is critical for interpreting complex datasets, yet the clarity and quality of visualizations vary widely across tools and applications. This study introduces a comprehensive framework for evaluating biomedical figures and benchmarking visualization platforms. We developed Metrics for Evaluation and Discretization of Biomedical Visuals using an Iterative Scoring algorithm (M.E.D.V.I.S.), a quantification system that systematically assesses figure quality based on four criteria: Complexity, color usage, whitespace, and the number of distinct visualizations. The algorithm integrates dimensionality reduction, clustering, and thresholding to classify figures and generate tailored feedback for improvement. In parallel, we conducted a comparative analysis of 26 widely used visualization tools, evaluating each based on ease of use, customizability, financial cost, and required background knowledge. To demonstrate real-world applicability, we present case studies on figure evaluation in published research and introduce SpatioView, an interactive, browser-based platform for exploring spatial omics data. Collectively, our findings highlight the need for standardized evaluation methods and provide accessible solutions for improving figure design in biomedical research, education, and industry.

Molecular network analysis offers powerful insights for plant improvement by capturing complex regulatory interactions. However, translating omics data across species presents significant challenges. Non-model crops such as soybean and lupin often lack comprehensive genomic resources, which complicates network analysis. Model species (e.g., Arabidopsis thaliana) provide rich data but may lack legume-specific pathways. This review synthesizes these challenges and examines legume networks in soybean, lupin, and the model legume, Medicago truncatula. Strategies such as multi-omics integration and Artificial Intelligence (AI)-driven tools, combined with wet lab validation studies such as clustered regularly interspaced short palindromic repeats (CRISPR), are discussed to bridge the gap between discovery and application. Ultimately, we conclude that cross-species multi-omics integration, empowered by AI and validated by gene editing, will be pivotal for translating network discoveries into resilient legume crops. Strategic investments in under-researched non-model legumes and advanced molecular tools are essential to ensure sustainable agriculture and future crop resilience.

Multimodal models have enabled the integration of digital pathology, radiology, clinical information, and omics data to enhance Colorectal cancer (CRC) care. This systematic review critically appraises Multimodal digital pathology techniques applied in CRC, their performance, and contrasts them with foundation models. We identified and screened 1601 studies published between January 2014 and August 2024 using PubMed, Web of Science, Scopus, and IEEE Xplore (PROSPERO protocol: 635831). The quality and bias of the 22 eligible studies were assessed using the Newcastle-Ottawa Scale. Our findings suggest that majority of the studies integrated different modalities to enhance diagnostic accuracy and survival prediction. Various fusion techniques have been used to extract novel features. Most studies did not undertake external validation. Compared to unimodal models, multimodal approaches demonstrate superior performance but challenges remain, including constructing multimodal datasets, managing data heterogeneity, ensuring temporal alignment, determining modality weighting, and improving interpretability.

Spatial transcriptomics technologies provide spatially-resolved measurements of gene expression through assays that can either target selected genes or capture transcriptome-wide expression profiles. The complexity and variability of these technologies and their associated data necessitate multi-step workflows integrating diverse computational methods and software packages. We provide a freely accessible, open-source, continuously updated and tested online book containing reproducible code examples, datasets, and discussion about data analysis workflows for spatial omics data using Bioconductor in R, including interoperability with Python.

Recent advances in high-throughput technologies have enabled observational studies to collect high-dimensional omic data. However, such data, often measured on small sample sizes, pose challenges to model-based clustering approaches such as Gaussian Mixture Models. Existing methods often fail to generalize due to model instability under complex mixture patterns. To overcome these limitations, we propose a natural-gradient variational inference framework for Gaussian mixture models named Praxis-BGM that incorporates informative priors—cluster-specific means, covariances, and structural connectivity—from large-scale reference data with known cluster or class labels to enable semi-supervised transfer learning. We derive natural-gradient updates that integrate prior knowledge, leveraging the Variational Online Newton algorithm. We also perform feature selection for clustering using Bayes Factors. Implemented using the JAX library for accelerator-oriented computation, Praxis-BGM is computationally efficient and scalable. We demonstrate the effectiveness of Praxis-BGM in extensive simulations and with two real-world applications: bulk transcriptomic datasets for breast cancer subtyping (the Cancer Genome Atlas Breast Invasive Carcinoma and the Molecular Taxonomy of Breast Cancer International Consortium), and transferring cell-type annotations between single-cell transcriptomic datasets produced by different single-cell RNA-seq technologies in a human pancreas study. Even when priors are partially mismatched with the target data, Praxis-BGM enhances semi-supervised clustering accuracy and biological interpretability.

Mediation analysis is a pivotal tool for elucidating the indirect effect of an environmental factor or treatment on disease through potentially high-dimensional omics data, such as gene expression profiles. However, traditional mediation analysis methods tailored for binary outcomes often rely on the rare disease assumption in logistic regression and provide inadequate measures of total mediation effect when multiple mediators have effects in different directions. In this paper, we develop a MEdiation analysis framework in LOgistic regression for high-Dimensional mediators and a binarY outcome (MELODY). It leverages a second-moment-based measure analogous to theR2for linear models to quantify the total mediation effect. We also develop a variable selection procedure for high-dimensional data to reduce bias introduced by non-mediators. Our comprehensive simulations demonstrate the superior performance of MELODY in scenarios with non-rare disease binary outcomes and high-dimensional mediators. We apply MELODY to the Framingham Heart Study of over 5000 individuals to analyze the mediation effects of metabolomics and transcriptomics data on the pathways from sex to incident coronary heart disease.

Genomic selection has become a widely adopted and effective breeding technology for modern genetic improvements in pigs. However, the core model currently used in genetic evaluation is primarily based on a linear mixed model, which accounts for only additive genetic effects. Non-additive effects and complex nonlinear interactions among genes or loci are often neglected, leaving substantial potential for improving the predictive ability of traits. To address this limitation, we here propose a Multi-omics Annotation and Residual Split strategy-based deep learning model (MARS). Through comprehensive comparisons and evaluations against various linear and nonlinear models across multiple pig traits, we demonstrate that the residual split indirect strategy effectively mitigates overfitting and underfitting issues commonly observed in deep learning models, thereby enhancing predictive accuracy for complex traits. Moreover, by incorporating multi-omics annotation information within a hierarchical feature selection procedure, our results show that it improves computational efficiency without significant sacrifices in prediction performance. It is foreseeable that our developed MARS would facilitate the application of artificial intelligence technology and the publicly available big omics data in the coming future of pig breeding.

Omics and Multiomics-Based Diagnostics for Invasive Candidiasis: Toward Precision Medicine.

ABSTRACTBackgroundThe 7‐Dehydrocholesterol reductase (DHCR7), a critical enzyme catalyzing the final step of the cholesterol biosynthesis pathway, has gained attention for its potential role in tumorigenesis. This study systematically investigated the association between DHCR7 expression and oncogenic processes across multiple cancer types.MethodsMulti‐omics data were obtained from The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) repositories. DHCR7 expression patterns were analyzed using Oncomine, TIMER, and GEPIA platforms. Prognostic significance was assessed via Kaplan–Meier plotter and GEPIA. Tumor stage correlations and immune/molecular subtype associations were evaluated using TISIDB. SangerBox facilitated analysis of DHCR7's associations with immune checkpoint (ICP) molecules, tumor mutational burden (TMB), microsatellite instability (MSI), mutant‐allele tumor heterogeneity (MATH), neoantigen load, and immune cell infiltration.ResultsDHCR7 exhibited significant overexpression in most malignancies, correlating with advanced tumor stage (p < 0.05), metastatic progression, and reduced overall survival (HR = 1.34, 95% CI: 1.18–1.52). Strong associations emerged between DHCR7 expression and critical immunomodulatory parameters: positive correlations with ICPs (PD‐L1: r = 0.62, CTLA4: r = 0.58). Significant links to TMB (p = 2.1e−5), MSI (p = 4.3e−4), and MATH (p = 7.8e−3). Distinct immune infiltration patterns, particularly in bladder carcinoma (BLCA), renal clear cell carcinoma (KIRC), and prostate adenocarcinoma (PRAD). Co‐expression network analysis revealed DHCR7's involvement in immune response regulation (GO:0002764, FDR = 0.003), leukocyte differentiation (GO:0002521, FDR = 0.012), and angiogenesis (GO:0001525, FDR = 0.018).ConclusionsThese pan‐cancer analyses identify DHCR7 as a multifaceted biomarker with dual prognostic and immunotherapeutic relevance. Its involvement in tumor immune microenvironment modulation suggests potential as a therapeutic target.

Omics analyses are crucial for understanding molecular mechanisms in biological research. The vast quantity of detected biomolecules presents a significant challenge in identifying potential biomarkers. Traditional methods rely on labor-intensive literature mining to extract meaningful insights from long lists of regulated candidates of biomolecules. To address this, we developed OmixLitMiner 2 (OLM2) to improve the efficiency of omics data interpretation, speed up the validation of results and accelerate further evaluation based on the selection of marker candidates for subsequent experiments. The updated tool utilizes UniProt for synonym and protein name retrieval and employs the PubMed database as well as PubTator 3.0 for searching titles or abstracts of available biomedical literature. It allows for advanced keyword-based searches and provides classification of proteins or genes with respect to their representation in the literature in relation to scientific questions. OLM2 offers improved functionality over the previous version and comes with a user-friendly Google Colab interface. In comparison to the previous version, OLM2 improves the retrieval of relevant publications and the classification of biomolecules. We use a case study of spatially resolved proteomic data from the mouse brain cortex to demonstrate that the tool significantly reduces the time compared to manual searches and enhances the interpretability of molecular analysis.

Substantiating chemical groups for read-across using molecular response profiles.

Mass spectrometry-based absolute quantitative proteomics of drug-metabolizing enzymes in human liver.