Carrero, Z. I. | Search bioRxiv

bioRxiv2026

Systematic functional drug testing in patient-derived models reveals ex vivo sensitivities associated with clinical outcome in rare solid tumors

Rare cancers are individually uncommon but collectively represent a substantial share of cancer burden, with limited systemic treatment options for many entities. Molecular profiling identifies targetable alterations, but actionable findings are limited and responses can vary despite a matched target. This motivates complementary approaches that directly assess tumor drug response. Here, we establish a biopsy-compatible ex vivo drug sensitivity testing platform optimized for low input and reproducibility. Patient-derived material was tested either directly or following ex vivo expansion. Functional profiling was performed within clinically relevant timelines across models from 126 patients with rare advanced solid tumors. Drug responses were consistent between model types. In most samples, we identified at least one potentially active compound, supporting feasibility at biopsy-scale. High in vitro sensitivity was associated with clinical benefit and progression-free survival. These findings support functional drug sensitivity testing as a complementary component in precision oncology for adults with rare cancers. Statement of SignificanceThis study presents a biopsy-compatible drug sensitivity testing platform for phenotype-based therapy stratification in rare cancers. It identifies actionable ex vivo drug responses and shows associations with clinical outcome in patients treated with screened therapies. These findings support functional testing as a complementary additional layer of stratification for therapeutic prioritization.

cancer biology↗

bioRxiv2025

IBDome: An integrated molecular, histopathological, and clinical atlas of inflammatory bowel diseases

Multi-omic and multimodal datasets with detailed clinical annotations offer significant potential to advance our understanding of inflammatory bowel diseases (IBD), refine diagnostics, and enable personalized therapeutic strategies. In this multi-cohort study, we performed an extensive multi-omic and multimodal analysis of 1,002 clinically annotated patients with IBD and non-IBD controls, incorporating whole-exome and RNA sequencing of normal and inflamed gut tissues, serum proteomics, and histopathological assessments from images of H&E-stained tissue sections. Transcriptomic profiles of normal and inflamed tissues revealed distinct site-specific inflammatory signatures in Crohns disease (CD) and ulcerative colitis (UC). Leveraging serum proteomics, we developed an inflammatory protein severity signature that reflects underlying intestinal molecular inflammation. Furthermore, foundation model-based deep learning accurately predicted histologic disease activity scores from images of H&E-stained intestinal tissue sections, offering a robust tool for clinical evaluation. Our integrative analysis highlights the potential of combining multi-omics and advanced computational approaches to improve our understanding and management of IBD.

molecular biology↗

bioRxiv2025

Deep Learning for Biomarker Discovery in Cancer Genomes

BackgroundAccurate determination of genomic biomarkers from tumor sequencing is fundamental to precision oncology, informing disease classification and treatment decisions. In practice, biomarker inference relies on computational pipelines that often compress high-dimensional mutation data into predefined summaries such as mutational signatures or composite genomic features. While robust and widely adopted, these representations may not fully capture the complexity of cancer genomes. Deep learning (DL) offers an end-to-end alternative by learning features directly from raw genomic data. However, clinical translation remains challenging due to limited empirical validation of new DL models and a lack of systematic comparisons with established machine learning (ML) baselines, particularly when transitioning from information-rich genome or exome data to real-world targeted sequencing profiles. Here, we compare state-of-the-art DL architectures with classical ML models across variant-level, copy-number (CNV), and multimodal inputs, using microsatellite instability (MSI) and homologous recombination deficiency (HRD) prediction as oncologically relevant tasks. We aim to derive practical guidance on modelling strategies across different data modalities and clinical sequencing contexts. MethodsFor MSI and HRD prediction, we trained multiple DL models, including supervised and self-supervised encoders, alongside feature-based ML approaches using tumor mutation data, copy-number alterations, and their multimodal combinations. Analyses were conducted on 5,647 patients in The Cancer Genome Atlas (TCGA), the Clinical Proteomic Tumor Analysis Consortium (CPTAC), and two targeted sequencing panel cohorts. Model performance was evaluated on both whole-exome and panel-based datasets, and explainability analysis were performed for both DL and ML models. ResultsFor MSI, DL demonstrated stronger generalization than ML on external validation data (F1 0.97 vs 0.76) and maintained comparatively high performance under pseudo-panels conditions, whereas ML performance dropped. In a real-world targeted panel cohort, DL again showed more robust generalization than ML, with performance partly affected by cross-assay variability. For HRD, incorporation of CNV data was the primary determinant of predictive performance. Once CNVs were included, DL and ML achieved similar accuracy on external datasets (F1 0.61 vs 0.58). In panel-based settings, DL retained an advantage over ML (F1 0.78 vs 0.62). Model interpretation analyses indicated that both DL and ML relied on mutation and chromosomal patterns consistent with established MSI and HRD biology. ConclusionOverall, predictive performance depended strongly on data availability and clinical sequencing context. When information-rich inputs were available, both DL and classical ML achieved robust biomarker prediction, with DL generally matching or exceeding ML performance. The most pronounced advantages of DL emerged in cross-assay evaluations and data-sparse settings, where generalization was more reliable. Notably, the best-performing DL models were lightweight and interpretable, supporting practical deployment. In clinical genomics workflows, such models may complement established pipelines by leveraging patient sequencing data to provide additional evidence for treatment-relevant biomarker assessment.

bioinformatics↗