Biology subjects

Rodriguez-Mier, P.

Publications and source records attributed to Rodriguez-Mier, P..

9 recordsLinked to original sources

bioRxiv2026

Towards Useful and Private Synthetic Omics: Community Benchmarking of Generative Models for Transcriptomics Data

BackgroundThe synthesis of anonymized data derived from real-world cohorts offers a promising strategy for regulatory-compliant and privacy-preserving biological data sharing, potentially facilitating model development that can improve predictive performance. However, the extent to which generative models can preserve biological signals while remaining resilient to adversarial privacy attacks in high-dimensional omics contexts remains underexplored. To address this gap, the CAMDA 2025 Health Privacy Challenge launched a community-driven effort to systematically benchmark synthetic and privacy-preserving data generation for bulk RNA-seq cohorts. ResultsBuilding on this initiative, we systematically benchmarked 11 generative methods across two cancer cohorts ([~]1,000 and [~]5,000 patients) over 978 landmark genes. Methods were evaluated across complementary axes of distributional fidelity, downstream utility, biological plausibility and empirical privacy risk, with emphasis on trade-offs between vulnerability to membership inference attacks (MIA) and other evaluation dimensions. Expressive deep generative models achieved strong predictive utility and differential expression recovery, but were often more vulnerable to membership inference risk. Differentially private methods improved resistance to attacks at the cost of reduced utility, while simpler statistical approaches offered competitive utility with moderate privacy risk and fast training. ConclusionsSynthetic bulk RNA-seq quality is inherently multi-dimensional and shaped by trade-offs between utility, biological preservation and privacy. Our results indicate that differences in model architecture drive distinct trade-offs across these axes, suggesting that model choice should align with dataset characteristics, intended downstream use and privacy requirements. Privacy risk should also be assessed using multiple complementary attack methods and, where possible, formal differential privacy protection.

bioinformatics↗

bioRxiv2025

ScAPE: A lightweight multitask learning baseline method to predict transcriptomic responses to perturbations

Predicting biological responses to perturbations is essential for advancing drug development efforts. While high-throughput technologies have greatly enhanced our ability to study perturbation effects at the molecular level, accurately modeling these responses remains challenging due to the complex underlying molecular interactions, the vast dimensionality of cellular response, the limited amount of data, and technical limitations in data collection. Machine learning (ML) approaches have emerged as powerful tools for addressing these challenges. However, the observation in recent benchmarks that complex models do not consistently outperform simpler ones highlights the need for robust and efficient baselines for evaluating new methods. We present ScAPE (Single Cell Analysis of Perturbational Effects), an ML method for predicting drug-induced gene expression changes. ScAPE leverages only aggregated gene-level statistics across cell types and perturbations as input features, yet achieves good performance despite this simplicity. Its effectiveness was demonstrated in the Single-cell Perturbation Prediction NeurIPS 2023 challenge, where it ranked among the top-performing methods. We describe the core model and introduce extensions that enhance its utility as a baseline, including a multitask learning extension for simultaneous prediction of multiple transcriptional response metrics. A software API is provided to support adoption and integration into research pipelines. Benchmarking against other leading challenge entries and general-purpose ML approaches, using the original challenge metric, supports ScAPE as a simple, lightweight baseline for evaluating methods that predict perturbation effects at the gene-expression level. ScAPE is built using Keras 3, enabling support for different backends such as TensorFlow, JAX, or PyTorch, and freely available as an open-source package at github.com/scapeML/scape.

bioinformatics↗

bioRxiv2024

NetworkCommons: bridging data, knowledge and methods to build and evaluate context-specific biological networks

SummaryWe present NetworkCommons, a platform for integrating prior knowledge, omics data, and network inference methods, facilitating their usage and evaluation. NetworkCommons aims to be an infrastructure for the network biology community that supports the development of better methods and benchmarks, by enhancing interoperability and integration. Availability and ImplementationNetworkCommons is implemented in Python and offers programmatic access to multiple omics datasets, network inference methods, and benchmarking setups. It is a free software, available at https://github.com/saezlab/networkcommons. Contactsaezlab@ebi.ac.uk, martin.garrido@embl.de Supplementary DataContribution guidelines, additional figures, and descriptions for data, knowledge, methods, evaluation strategies and their implementation are available in the Supplementary Data and in the NetworkCommons documentation at https://networkcommons.readthedocs.io/.

bioinformatics↗

bioRxiv2024

Unified knowledge-driven network inference from omics data

Analysing omics data requires computational methods to effectively handle its complexity and to derive meaningful hypotheses about molecular mechanisms. While data-driven statistical and machine learning methods can identify patterns from omics data across multiple samples, they typically require a large number of samples and they often lack interpretability and alignment with existing biological knowledge. In contrast, knowledge-based network methods integrate molecular data with prior knowledge to provide results that are biologically interpretable, but they lack both a unified mathematical framework, leading to ad-hoc solutions specific to particular data types or prior knowledge, limiting their generalisability, and a common modelling interface for programmatic manipulation, restricting method extensions. Furthermore, existing methods generally cannot perform joint network inference across multiple samples or conditions, which restricts their capacity to capture shared mechanisms, making these methods more sensitive to noise and prone to overfitting. To address these limitations, we introduce CORNETO (Constrained Optimisation for the Recovery of NETworks from Omics), a unified framework for knowledge-driven network inference. CORNETO redefines the joint inference task as a constrained optimisation problem with a penalty that induces structured sparsity, allowing for simultaneous network inference across multiple samples. The framework is highly flexible and supports a wide variety of prior knowledge networks--undirected, directed and signed graphs, as well as hypergraphs--enabling the generalisation and improvement of many network inference methods, despite their seemingly different assumptions. We demonstrate its utility by presenting novel extensions of methods for signalling, metabolism and protein-protein interactions. We show how these new methods improve the performance of traditional techniques on a diverse set of biological tasks using simulated and real data. CORNETO is available as an open-source Python package (github.com/saezlab/corneto), facilitating researchers in extending, reusing, and harmonising methods for network inference.

bioinformatics↗

bioRxiv2024

Evaluating signaling pathway inference from kinase-substrate interactions and phosphoproteomics data

Cellular signaling plays a vital role in how cells communicate and adapt to both environmental and internal cues. At the molecular level, signaling is largely driven by phosphorylation cascades controlled by kinases. Because of this, kinase-driven signaling pathways are used as a conceptual framework to interpret molecular data across biological contexts. However, signaling pathways were created using limited throughput technologies. As knowledge of kinase-substrate interactions grows through novel computational and experimental approaches, and phosphoproteomic methods improve their coverage and accuracy, traditional signaling pathways need to be revisited. In this study, we critically assess context-specific signaling pathway reconstruction using phosphoproteomics and kinase-substrate networks. We first integrate literature, protein language models, and peptide array data to create a state-of-the-art kinase-substrate network. Focusing on epidermal growth factor (EGF), we conduct a meta-analysis of recent short-term response phosphoproteomics studies, which we complement with three own datasets, representing the most comprehensive characterization of the EGF response available to date. Using three alternative computational methods, we infer kinase-driven pathways, which we compare to multiple ground truth sets, including the canonical pathway, experimentally validated interactions, and correlation supported interactions. Our findings reveal that literature-curated networks, when combined with network propagation, yield the best recovery of ground truth interactions. We found that up to 90% of data-supported direct interactions are absent from current ground truth sets, indicating many unexplored, but data supported kinase interactions. Our results challenge traditional views on signaling pathways and illustrate how to develop new mechanistic hypotheses using phosphoproteomics and network methods.

systems biology↗

bioRxiv2023

Genome scale metabolic network modelling for metabolic profile predictions

Metabolic profiling (metabolomics) aims at measuring small molecules (metabolites) in complex samples like blood or urine for human health studies. While biomarker-based assessment often relies on a single molecule, metabolic profiling combines several metabolites to create a more complex and more specific fingerprint of the disease. However, in contrast to genomics, there is no unique metabolomics setup able to measure the entire metabolome. This challenge leads to tedious and resource consuming preliminary studies to be able to design the right metabolomics experiment. In that context, computer assisted metabolic profiling can be of strong added value to design more quickly and efficiently metabolomics studies. We propose a constraint-based modelling approach which predicts in silico profiles of metabolites that are more likely to be differentially abundant under a given metabolic perturbation (e.g. due to a genetic disease), using flux simulation. In genome-scale metabolic networks, the fluxes of exchange reactions, also known as the flow of metabolites through their external transport reactions, can be simulated and compared between control and disease conditions in order to calculate changes in metabolite import and export. These import/export flux differences would be expected to induce changes in circulating biofluid levels of those metabolites, which can then be interpreted as potential biomarkers or metabolites of interest. In this study, we present SAMBA (SAMpling Biomarker Analysis), an approach which simulates fluxes in exchange reactions following a metabolic perturbation using random sampling, compares the simulated flux distributions between the baseline and modulated conditions, and ranks predicted differentially exchanged metabolites as potential biomarkers for the perturbation. We show that there is a good fit between simulated metabolic exchange profiles and experimental differential metabolites detected in plasma, such as patient data from the disease database OMIM, and metabolic trait-SNP associations found in mGWAS studies. These biomarker recommendations can provide insight into the underlying mechanism or metabolic pathway perturbation lying behind observed metabolite differential abundances, and suggest new metabolites as potential avenues for further experimental analyses. Author summaryAssociating diseases and other metabolic disruptions with physiological markers is key for diagnostic and personalised medicine. These markers can be metabolites - small molecules involved in every living beings metabolism, and can be measured in biofluids such as blood or urine using metabolic profiling (metabolomics). Nevertheless, this experimental metabolomics design needs to be tailor made for each disease to ensure that most relevant metabolites will be detected. The selection of metabolites to analyse for future experiments can be time-consuming and expensive. In this paper, we build upon an existing computational method for simulating metabolite changes in a human model. This provides a prediction of the change in biofluid abundance of every known metabolite involved in human metabolism in a potentially large number of metabolic situations. The newly introduced method produces a change score and a rank for each metabolite in each condition. We show the strong potential of the approach by comparing predictions with experimental results.

systems biology↗

bioRxiv2022

Suggesting disease associations for overlooked metabolites using literature from metabolic neighbours

In human health research, metabolic signatures extracted from metabolomics data are a strong-added value for stratifying patients and identifying biomarkers. Nevertheless, one of the main challenges is to interpret and relate these lists of discriminant metabolites to pathological mechanisms. This task requires experts to combine their knowledge with information extracted from databases and the scientific literature. However, we show that a large fraction of metabolites are rarely or never mentioned in the literature. Consequently, these overlooked metabolites are often set aside and the interpretation of metabolic signatures is restricted to a subset of the significant metabolites. To suggest potential pathological phenotypes related to these understudied metabolites, we extend the guilt by association principle to literature information by using a Bayesian framework. With this approach, we suggest more than 35,000 associations between 1,047 overlooked metabolites and 3,288 diseases (or disease families). All these newly inferred associations are freely available on the FORUM ftp server (See information at https://github.com/eMetaboHUB/Forum-LiteraturePropagation.).

bioinformatics↗

bioRxiv2021

Pathway analysis in metabolomics: pitfalls and best practice for the use of over-representation analysis

Over-representation analysis (ORA) is one of the commonest pathway analysis approaches used for the functional interpretation of metabolomics datasets. Despite the widespread use of ORA in metabolomics, the community lacks guidelines detailing its best-practice use. Many factors have a pronounced impact on the results, but to date their effects have received little systematic attention in the field. We developed in-silico simulations using five publicly available datasets and illustrated that changes in parameters, such as the background set, differential metabolite selection methods, and pathway database choice, could all lead to profoundly different ORA results. The use of a non-assay-specific background set, for example, resulted in large numbers of false-positive pathways. Pathway database choice, evaluated using three of the most popular metabolic pathway databases: KEGG, Reactome, and BioCyc, led to vastly different results in both the number and function of significantly enriched pathways. Metabolomics data specific factors, such as reliability of compound identification and assay chemical bias also impacted ORA results. Simulated metabolite misidentification rates as low as 4% resulted in both gain of false-positive pathways and loss of truly significant pathways across all datasets. Our results have several practical implications for ORA users, as well as those using alternative pathway analysis methods. We offer a set of recommendations for the use of ORA in metabolomics, alongside a set of minimal reporting guidelines, as a first step towards the standardisation of pathway analysis in metabolomics. Author summaryMetabolomics is a rapidly growing field of study involving the profiling of small molecules within an organism. It allows researchers to understand the effects of biological status (such as health or disease) on cellular biochemistry, and has wide-ranging applications, from biomarker discovery and personalised medicine in healthcare to crop protection and food security in agriculture. Pathway analysis helps to understand which biological pathways, representing collections of molecules performing a particular function, are involved in response to a disease phenotype, or drug treatment, for example. Over-representation analysis (ORA) is perhaps the most common pathway analysis method used in the metabolomics community. However, ORA can give drastically different results depending on the input data and parameters used. In this work, we have established the effects of these factors on ORA results using computational simulations applied to five real-world datasets. Based on our results, we offer the research community a set of best-practice recommendations applicable not only to ORA but also to other pathway analysis methods to help ensure the reliability and reproducibility of results.

bioinformatics↗

bioRxiv2020

Diversity-based enumeration of optimal context-specific metabolic networks

The correct identification of metabolic activity in tissues or cells under different environmental or genetic conditions can be extremely elusive due to mechanisms such as post-transcriptional modification of enzymes or different rates in protein degradation, making difficult to perform predictions on the basis of gene expression alone. Context-specific metabolic network reconstruction can overcome these limitations by leveraging the integration of multi-omics data into genome-scale metabolic networks (GSMN). Using the experimental information, context-specific models are reconstructed by extracting from the GSMN the sub-network most consistent with the data, subject to biochemical constraints. One advantage is that these context-specific models have more predictive power since they are tailored to the specific organism and condition, containing only the reactions predicted to be active in such context. A major limitation of this approach is that the available information does not generally allow for an unambiguous characterization of the corresponding optimal metabolic sub-network, i.e., there are usually many different sub-network that optimally fit the experimental data. This set of optimal networks represent alternative explanations of the possible metabolic state. Ignoring the set of possible solutions reduces the ability to obtain relevant information about the metabolism and may bias the interpretation of the true metabolic state. In this work, we formalize the problem of enumeration of optimal metabolic networks, we implement a set of techniques that can be used to enumerate optimal networks, and we introduce DEXOM, a novel strategy for diversity-based extraction of optimal metabolic networks. Instead of enumerating the whole space of optimal metabolic networks, which can be computationally intractable, DEXOM samples solutions from the set of optimal metabolic sub-networks maximizing diversity in order to obtain a good representation of the possible metabolic state. We evaluate the solution diversity of the different techniques using simulated and real datasets, and we show how this method can be used to improve in-silico gene essentiality predictions in Saccharomyces Cerevisiae using diversity-based metabolic network ensembles. Both the code and the data used for this research are publicly available on GitHub1.

bioinformatics↗