Biology subjects

Novella Rausell, C.

Publications and source records attributed to Novella Rausell, C..

4 recordsLinked to original sources

bioRxiv2026

Zero-Shot Metabolite Prediction from Gene Expression via Physics-Informed Graph Neural Networks

Predicting metabolite concentrations from gene expression is instrumental for linking regulatory programs to metabolic phenotypes. Prior approaches rely on static enzyme-metabolite mappings and often omit data-driven learning or biochemical constraints, limiting their ability to generalize to new metabolites. We present GAZE (Graph Attention for Zero-shot metabolite Estimation), a physics-informed graph neural network that integrates enzyme expression, Enzyme Commission functional embeddings, and ChemBERTa metabolite descriptors within a unified metabolic graph (5,414 nodes, 16,307 edges). A Metabolite-Conditioned Reader uses each metabolites SMILES embedding to query learned pathway representations, enabling zero-shot prediction with no metabolite-specific parameters. We evaluate GAZE in three scenarios: (i) standard cross-validation on the Cancer Atlas of Metabolic Profiles (18,044 genes, 180 metabolites, 867 cell lines), achieving R2 = 0.816; (ii) leave-one-metabolite-out (LOMO) zero-shot evaluation across 50 held-out metabolites, where the physics-informed variant halves the median R2 deficit relative to a standard GNN baseline (-0.34 vs. -0.72), with 30% of unseen metabolites achieving positive R2; and (iii) external validation on an independent clear cell renal cell carcinoma tissue cohort (220 samples), where GAZE achieves median Spearman{rho} = 0.330 across 214 metabolites without fine-tuning. GAZE outperforms scCellFie, MEBOCOST, and UnitedMet across all evaluation settings.

bioinformatics↗

bioRxiv2025

Image-guided alignment of consecutive multi-modal tissue slides

Multi-modal spatial data analysis often requires precise physical alignment of consecutive tissue sections, a process that can be challenging and typically relies on shared molecular markers or image recognition techniques. Here, we introduce COAST (Consecutive multi-Omics Alignment of Spatial Tissues), a method to reliably physically align consecutive tissue sections to produce a unified multi-modal molecular dataset suitable for downstream applications. COAST relies exclusively on the images associated with spatial data, eliminating the need for common molecular features or prior annotations. We demonstrate the effectiveness of COAST using spatial transcriptomics slides, where it achieves performance comparable to established uni-modal alignment tools. Applying COAST to spatial transcriptomics and metabolomics/lipidomics tissue sections from a mouse model of ischemia reperfusion injury allowed the investigation of lipid/metabolite features of transcriptionally-defined cell types. Overall, COAST offers a streamlined and integrative solution for multi-modal spatial data alignment.

bioinformatics↗

bioRxiv2023

Integrative learning of disentangled representations from single-cell RNA-sequencing datasets

AO_SCPLOWBSTRACTC_SCPLOWSingle-cell RNA-sequencing is instrumental in studying cellular diversity in biological systems. Using batch correction methods, cell identities are often jointly defined across multiple conditions, individuals, or modalities. These approaches overlook group-specific information and require either paired data or matching features across datasets. Here we present shared-private Variational Inference via Product of Experts with Supervision (spVIPES), a framework to analyze the shared and private components of unpaired groups of cells with non-matching features. spVIPES represents the cells from the different groups as a composite of private and shared factors of variation using a probabilistic latent variable model. We evaluate the performance of spVIPES with a simulated dataset and apply our model in three different scenarios: (i) cross-species comparisons, (ii) regeneration following long and short acute kidney injury, and (iii) IFN-{beta} stimulation of PMBCs. In our study, we demonstrate that spVIPES accurately disentangles distinct sources of variation into private and shared representations while matching current state-of-the-art methods for batch correction. Furthermore, spVIPES shared space outperforms alternatives models at learning cell identities across datasets with non-matching features. We implemented spVIPES using the scvi-tools framework and release it as an open-source software at https://github.com/nrclaudio/spVIPES.

bioinformatics↗

bioRxiv2022

A comprehensive mouse kidney atlas enables rare cell population characterization and robust marker discovery

The cellular diversity and complexity of the kidney are on par with its physiological intricacy. Although our anatomical understanding of the different segments and their functions is supported by a plethora of research, the identification of distinct and rare cell populations and their markers remains elusive. Here, we leverage the large number of cells and nuclei profiles using single-cell (scRNA-seq) and single-nuclei (snRNA-seq) RNA-sequencing to build a comprehensive atlas of the adult mouse kidney. We created MKA (Mouse Kidney Atlas) by integrating 59 publicly available single-cell and single-nuclei transcriptomic datasets from eight independent studies. The atlas contains more than 140.000 cells and nuclei covering different single-cell technologies, age, and tissue sections. To harmonize annotations across datasets, we constructed a hierarchical model of the cell populations present in our atlas. Using this hierarchy, we trained a model to automatically identify cells in unannotated datasets and evaluated its performance against well-established methods and annotation references. Our learnt model is dynamic, allowing the incorporation of novel cell populations and refinement of known profiles as more datasets become available. Using MKA and the learned model of cellular hierarchies, we predicted previously missing cell annotations from several studies and characterized well-studied and rare cell populations. This allowed us to identify reproducible markers across studies for poorly understood cell types and transitional states.

bioinformatics↗