The Transcriptomic and Proteomic Molecular Signatures of Equine Multiple-System Neuropathy (Grass Sickness).

Machine learning is reshaping bioinformatics and biomedicine by enabling the analysis of high-dimensional molecular, clinical, imaging, and network-based data. With the rapid accumulation of genomic, transcriptomic, proteomic, metabolomic, imaging, and electronic health-related datasets, machine learning has become an important approach for disease biomarker discovery, clinical risk prediction, multi-omics integration, drug target identification, biomedical image analysis, and network-based biological interpretation [,,,]. In parallel, recent advances in biological artificial intelligence, including foundation and generalist models, are further expanding the scope of computational biology from task-specific prediction toward broader representation learning and cross-domain biological reasoning []. This Special Issue, “Machine Learning Applications in Bioinformatics and Biomedicine: 3rd Edition,” comprises ten contributions that reflect recent progress in this rapidly developing field. The published papers cover a broad range of biomedical scenarios, including cancer diagnosis and prognosis, ovarian cancer metabolomics, chronic inflammation-associated cancer susceptibility, Alzheimer’s disease pathology, respiratory disease diagnosis, schizophrenia risk stratification, rare disease genotype–phenotype analysis, and network pharmacology. Collectively, these contributions highlight advances in algorithm development, feature selection, interpretable modeling, multimodal data integration, and automated biomedical analysis platforms. This editorial summarizes the core content and major scientific contributions of these studies from the perspective of the Guest Editor. The identification of disease biomarkers from genomic and transcriptomic data remains one of the most active areas in biomedical machine learning. Several of the contributions focused on cancer classification, gene selection, and prognostic modeling. Ghuriani et al. (Contribution 1) proposed XGB-BIF, an XGBoost-driven biomarker identification framework for detecting gastric, breast, and lung cancers using human genomic data. By combining XGBoost-based feature selection with classical machine learning classifiers, including support vector machines, logistic regression, and random forest, the framework achieved strong classification performance and identified cancer-associated genes with potential diagnostic relevance. The integration of SHAP, LIME, pathway enrichment, and survival analysis further enhanced the interpretability and translational value of the identified biomarkers. Alkamli and Alshamlan (Contribution 2) introduced GNR, a genetic-embedded nuclear reaction optimization algorithm combined with an F-score filter for gene selection in cancer classification. By improving the exploration and exploitation capability of nuclear reaction optimization through a genetic uniform crossover mechanism, GNR selected compact gene subsets and achieved high classification accuracy across multiple microarray cancer datasets. Zhou et al. (Contribution 3) developed a glycosyltransferase-related prognostic model for clear cell renal cell carcinoma. Through the systematic comparison of 117 machine learning algorithm combinations, the authors established the Glycosyltransferases Risk Score model, which stratified patients into distinct prognostic groups and provided insights into tumor mutation burden, immune microenvironment, immunotherapy response, and key genes such as TYMP and GCNT4. Together, these studies demonstrate the value of machine learning in extracting informative molecular signatures from high-dimensional cancer data and supporting precision oncology.