Biology subjects

Matos-Romero, V.

Publications and source records attributed to Matos-Romero, V..

3 recordsLinked to original sources

bioRxiv2026

Quantitative Donor Histology and Transcriptomics Link Pancreatic Composition and Precancers Across the Adult Lifespan

Background & Aims: Pancreatic intraepithelial neoplasia (PanIN) is the primary precursor to invasive pancreatic ductal adenocarcinoma (PDAC). While PanIN is common in the general population, PDAC is rare. PDAC risk rises with age, diabetes mellitus, and obesity, but whether this increased risk reflects a greater precancer burden is unknown, as previous studies assessed PanIN presence and grade. We aimed to quantify pancreatic composition and PanIN burden across the adult lifespan to identify demographics of donors carrying disproportionate precancer burden. Methods: We applied CODAvision, a deep-learning segmentation pipeline, to 410 high-quality pancreas sections from Genotype-Tissue Expression (GTEx) donors aged 21 to 70 years. We screened twenty-seven donor phenotypes against 14 compositional measures, applying Benjamini-Hochberg correction for multiple testing. We further related tissue composition and donor phenotype to bulk gene expression in 281 donors to understand which demographics are associated with a compositional or transcriptional fingerprint. Results: PanIN was present in 83% of donors. Acinar proportion declined with age, offset by fat and stroma. Median PanIN burden did not change with age, though the fitted 90th percentile rose from 0.21% to 0.72% of exocrine tissue (uncorrected P=.002). The 39 sections (10%) with high-outlier PanIN burden came disproportionately from donors with type 2 diabetes mellitus and hypertension and had greater stromal proportion and higher fibroblast and immune-cell densities. Both cell types were denser in peri-PanIN stroma than in the surrounding pancreatic parenchyma. PanIN abundance was associated with a gastric transcriptional signature including VSIG1, TFF1, and MUC5AC. Conclusions: Most donors carried little PanIN at any age, whereas high burden was concentrated in a minority of donors more likely to have type 2 diabetes mellitus, hypertension, and inflamed, fibrotic stroma. Because PanIN was present in most donors, its presence did not distinguish high-risk groups, whereas quantitative burden did, concentrating on the metabolic phenotypes already under consideration for pancreatic surveillance.

bioinformatics↗

bioRxiv2026

TiSAM: fusing spatial and histopathology foundation models for whole-slide image segmentation

Semantic segmentation of whole-slide images (WSIs) quantifies tissue composition and architecture, but existing models require extensive annotation and lose accuracy as the number of tissue classes increases or on heterogeneous, cross-scanner datasets. Here, we present the Tissue Segment Anything Model (TiSAM), which fuses SAM3 spatial features with UNI2-h pathology features within a multi-scale decoder. Because public benchmarks contain few tissue classes, we release MegaMouse, comprising 36,833 annotations of 52 tissue classes from 12 mouse organs, and an 11-class human pancreas dataset. TiSAM achieved the highest overall performance across four public datasets compared to existing convolutional neural network and vision transformer models, maintained high accuracy on MegaMouse, and maintained high accuracy in the human pancreas dataset when trained on as few as 5 annotated WSIs where previous integrations of SAM with pathology foundation models did not. TiSAM segmentations were also more consistent across eight slide scanners compared to other models. These results establish TiSAM as a general and highly adaptable framework for tissue segmentation, enabling studies containing many tissue classes, limited annotations, or images from multiple scanners.

bioinformatics↗

bioRxiv2025

CODAvision: best practices and a user-friendly interface for rapid, customizable segmentation of medical images

Image-based machine learning tools have emerged as powerful resources for analyzing medical images, with deep learning-based semantic segmentation commonly utilized to enable spatial quantification of structures in images. However, customization and training of segmentation algorithms requires advanced programming skills and intricate workflows, limiting their accessibility to many investigators. Here, we present a protocol and software for automatic segmentation of medical images guided by a graphical user interface (GUI) using the CODAvision algorithm. This workflow simplifies the process of semantic segmentation of microanatomical structures by enabling users to train highly customizable deep learning models without extensive coding expertise. The protocol outlines best practices for creating robust training datasets, configuring model parameters, and optimizing performance across diverse biomedical image modalities. CODAvision enhances the usability of the CODA algorithm (Nature Methods, 2022) by streamlining parameter configuration, model training, and performance evaluation, automatically generating quantitative results and comprehensive reports. We expand beyond the original implementation of CODA to serial histology by demonstrating robust performance across numerous medical image modalities and diverse biological questions. We provide sample results in data types including histology, magnetic resonance imaging (MRI), and computed tomography (CT). We demonstrate the diverse use of this tool in applications including quantification of metastatic burden in in vivo models and deconvolution of spot-based spatial transcriptomics datasets. This protocol is designed for researchers with interest in rapid design of highly customizable semantic segmentation algorithms and a basic understanding of programming and anatomy.

bioinformatics↗