From Biomarkers to Biological Insight: Explainable AI in Genetics and Genomics

High-throughput sequencing, single-cell technologies, and large-scale population studies have transformed genetics and genomics into data-intensive sciences. While these advances enable comprehensive molecular profiling across genomes, epigenomes, transcriptomes, and other omics layers, translating high-dimensional molecular data into mechanistic biological insight remains a central challenge. Classical statistics, including linear regression, GWAS, and standard dimensionality reduction approaches, have provided foundational discoveries, yet they rely on assumptions of additivity and limited interaction structures. Such assumptions often constrain their ability to capture non-linear effects, regulatory dependencies, and cross-omics interactions that underline complex traits and biological systems. Artificial intelligence (AI) methods offer the expressive capacity to model these high-dimensional and heterogeneous relationships without pre-specifying interaction structures. They have demonstrated strong performance in tasks such as biomarker discovery, variant prioritisation, gene and regulatory annotation, and prediction of complex phenotypes. However, as these AI models increasingly inform experimental prioritisation and, in some contexts, translational decision-making, a critical limitation becomes evident: predictive performance alone does not equate to biological understanding. In genomics, models are not used solely for classification or risk prediction, they are frequently expected to generate hypotheses, reveal regulatory mechanisms, and support mechanistic interpretation of genotype-phenotype relationships. Without explicit attribution of predictions to specific genes, variants, regulatory regions, or molecular layers, AI risks remaining a black box rather than a discovery framework. By enabling structured attribution, interaction analysis, and biologically grounded interpretation, the explainable AI (XAI) transforms predictive models into tools for insight generation, reproducibility, hypothesis-driven research and knowledge discovery. XAI is therefore not merely a transparency add-on, but a methodological necessity to bridge computational modelling and biological inference. The purpose of this research topic focuses on XAI approaches applied to high-dimensional genetic and genomic data, where explainability directly supports mechanistic insight. Particular emphasis is placed on genotype-phenotype modelling, regulatory genomics, functional interpretation of coding and non-coding variation, and multi-omics integration. While the core scientific questions are shared across all domains of life, we also welcome contributions from plant and agricultural genomics, microbiome and metagenomics, and environmental or evolutionary genomics, provided that explainability plays a central role in enabling biological interpretation. As well as case studies that showcase the importance of explainability in AI applications used to obtain biological conclusions, guide biomarker identification, and illuminate genotype-phenotype relationships. Our goal is to provide a home for work that moves from prediction toward insight, fostering dialogue between AI researchers and the genetics and genomics communities.We invite original research articles, and reviews that address any of these specific themes:• Explainable AI approaches that uncover non-linear interactions, and cross-omics dependencies in genotype-phenotype modelling.• Explainable AI for biomarker discovery, variant prioritisation in high-dimensional genomic datasets.• Explainable models for gene regulation, functional genomics, and sequence-based prediction.• Representation learning strategies explicitly designed to preserve biological explainability.• Multi-omics integration frameworks where explainability supports mechanistic understanding rather than predictive performance alone. • Applications of XAI in plant and agricultural genomics, microbiome and metagenomics, and environmental or evolutionary genomics, where explainability directly informs biological or translational insight.• Comparative studies and practical perspectives on explainability tools in genetic applications.• Reviews on current challenges and limitations, best practices, and future directions for XAI in genetics and genomics.• Conceptual frameworks positioning XAI as a bridge between predictive modelling and biological discovery. Topic editors: Pierfrancesco Novielli, Mariangela Santorsola Source: https://www.frontiersin.org/research-topics/77942/from-biomarkers-to-biological-insight-explainable-ai-in-genetics-and-genomics