Understanding Synergistic Interactions among Pathology Foundation Models via Adaptive Fusion

End-to-end multimodal pathology foundation model with clinical dialogue.

PRISM2 demonstrates how language-supervised pretraining provides a scalable, clinically grounded signal for generalizable pathology representations, bridging human diagnostic reasoning and foundation model performance.

Biologically Informed Deep Neural Networks for Multi-Omic Integration, Pathway Activity Inference and Risk Stratification in Cancer

This work reports Pathway Activity Autoencoders for the multi-omics setting, which embed prior knowledge via pathway-informed architectural constraints, fostering interpretability, while preserving representational power, in the context of breast cancer.

Robust and Interpretable Metagenomic Modeling Through Structure-Aware Multi-View Learning and Attribution-Guided Biological Insight

Integrative modeling of metagenomic and clinical data can advance the study of host phenotypes, but remains challenged by cross-view heterogeneity, uncertain generalizability, and poor interpretability. We developed SAMECAT (Structure-Aware Metagenomics multi-viEw Contrastive AlignmenT), a structure-aware deep learning framework that integrates species-level shotgun metagenomic profiles with mixed-type clinical covariates through view-specific encoders, clustering-informed contrastive alignment, and adaptive representation fusion. Using two independent Louisiana Osteoporosis Study datasets generated through distinct sequencing and bioinformatics pipelines (development n = 1,990; external evaluation n = 481), we evaluated SAMECAT for bone mineral density prediction at four skeletal sites. SAMECAT consistently outperformed single-view models, naive concatenation, alternative deep learning integration approaches, and established machine learning baselines, with performance gains largely preserved in cross-pipeline external evaluation. To improve biological interpretability, we developed a stability-oriented interpretation workflow that aggregates individually low-magnitude and diffusely distributed feature attributions into structured modules, revealing reproducible site-dependent patterns, coherent functional themes, and representative hub taxa. SAMECAT thus provides a robust and interpretable framework for multi-view metagenomic modeling of microbiome-associated host phenotypes.

OmicFormer: a statistical priors-informed transformer for accurate and generalizable omics prediction of diseases and complex traits

SAGE: Semantic Explainability of Attention-Based Survival Models in Computational Pathology

Attention-based multiple instance learning (ABMIL) is the predominant approach for slide-level prediction in computational pathology, yet its attention maps provide only local explanations: they indicate where a model focuses but not which histological features drive its predictions or how the model behaves across a patient cohort. We present Semantic Attention Global Explanations (SAGE), a post-hoc framework that extracts global, language-grounded explanations from a frozen ABMIL model. Using a pathology vision-language model, SAGE scores image patches against a dictionary of 25 histological concepts, aggregates these scores according to the model's learned attention, and quantifies how each concept relates to prediction risk across a cohort. Applied to survival prediction using seven TCGA cancer cohorts and three foundation models, SAGE recovered established prognostic features, such as the adverse association of necrosis, while revealing cancer-specific biology, including a favorable angiogenic signature in renal cell carcinoma consistent with known molecular subtypes. Ablation studies demonstrated that these associations depend on the model's learned attention rather than concept prevalence alone, and that the concept dictionary captures much of the prognostic information encoded by the foundation model features. Through semantically-grounded explanations, SAGE provides a scalable, model-agnostic framework for understanding what ABMIL survival models learn, enabling pathologists to interpret model behavior at the cohort level and offering the potential for biomarker identification.

The Virtual Tissues foundation model resolves spatial proteomics across scales.

Spatial proteomics technologies have transformed our understanding of complex tissue architecture in cancer but present unique challenges for computational analysis1. Each study uses a different marker panel and protocol, and most methods are tailored to single cohorts, which limits knowledge transfer and robust biomarker discovery. Here we present Virtual Tissues (VirTues), a general-purpose foundation model for spatial proteomics that learns marker-aware, multi-scale representations of proteins, cells, niches and tissues directly from multiplex imaging data. From a single pretrained backbone, VirTues supports marker reconstruction, cell segmentation and typing, niche annotation, spatial biomarker discovery and patient stratification, including zero-shot annotation across heterogeneous panels and datasets. In triple-negative breast cancer, VirTues-derived biomarkers predict anti-PD-L1 chemo-immunotherapy response2 and stratify disease-free survival in an independent cohort3, outperforming state-of-the-art biomarkers derived from the same datasets and current clinical stratification schemes.