Biology subjects

Poyatos, J. F.

Publications and source records attributed to Poyatos, J. F..

13 recordsLinked to original sources

bioRxiv2026

Development positions malignant cellular states but does notexplain their diversification

Epithelial cancers are often described as aberrant reactivations of embryonic or tissue-forming programs, but whether malignant cellular-state diversification is actually constrained by developmental trajectories remains unclear. Here, we present a quantitative framework to test this idea in pancreatic ductal adenocarcinoma (PDAC). Using representation learning on large-scale single-cell data, we build a reference space that captures the main axes of normal foregut and pancreatic epithelial variation. Malignant cells can be mapped into this space, showing that developmental biology helps interpret their identity. However, these coordinates account for only part of the variation. A substantial portion lay outside the reference space, and the residual component was structured rather than diffuse. Furthermore, displacements between cancer states showed weak alignment with canonical axes and did not correspond to simple progression along them, even when they retained measurable components within the broader subspace. Thus, developmental programs offer a useful coordinate system for describing where PDAC states sit relative to normal epithelia, but they do not explain the directions along which malignant states differ. Instead, malignant diversification follows additional structured, cancer-adaptive axes outside the dominant geometry of normal development. More generally, these results demonstrate how disease-state variation can be decomposed into shared coordinates of normal identity and distinct directions of pathological reprogramming. Significance StatementTumors are often said to reuse genetic programs from early tissue formation, but similarity to those programs does not mean cancer cells follow the same trajectories. We developed a quantitative framework to separate these ideas in pancreatic ductal adenocarcinoma. Variation tied to normal epithelial differentiation provided a useful coordinate system for locating malignant cell states, yet differences between those states showed weak alignment with canonical differentiation directions and included structured, cancer-adaptive programs. Thus, pancreatic cancer retains aspects of normal cell identity while diversifying along additional malignant axes. This distinction clarifies how insights from normal tissue programs should be applied to cancer plasticity and provides a general method for comparing disease states with normal cellular programs.

systems biology↗

bioRxiv2025

Redefining Housekeeping Genes in the Context of Vertebrate Development

Housekeeping genes are typically defined as genes active in all adult tissues and essential for basic cellular functions. However, this view overlooks genes that are crucial during early development. Using zebrafish as a model system, we examined gene expression from the first hours of embryogenesis through adulthood. We found that many genes expressed throughout development remain broadly expressed in adults, but others do not; and some adult housekeeping genes are not active early on. This reveals that essential gene functions shift across life stages. Genes active throughout development are more likely to be evolutionarily ancient, shared with humans, and cause defects in many organs when disrupted. Yet some genes expressed only at specific developmental stages, especially around gastrulation, can be equally essential. This supports the developmental hourglass model, where mid-embryogenesis is the most conserved stage. Our results broaden the concept of housekeeping genes to include genes essential across time, not only across adult tissues.

systems biology↗

bioRxiv2025

Capturing Context in Organismal Development with a Large Language Model

Understanding how regulatory complexity and constraint shape organismal development remains a central challenge in biology. The developmental hourglass framework posits that mid-embryogenesis -the phylotypic stage- is a period of heightened conservation and coordinated regulatory organization. We test this hypothesis using Zebraformer, a transformer-based language model trained on single-cell transcriptomic data from zebrafish embryos. Zebraformer learns context-sensitive representations that capture temporal progression, anatomical identity, and regulatory relationships, yielding gene and cell embeddings that recapitulate the developmental axis and increasing transcriptional divergence over time. In contrast, attention-derived gene networks reveal a transient reorganization of regulatory architecture during the phylotypic stage, marked by tightly coordinated gene modules, reduced cross-module connectivity, and diminished local redundancy. Sensitivity to perturbation emerges specifically when regulatory interaction structure is taken into account, rather than from perturbation magnitude alone, highlighting that constraint during this stage is embedded in network topology rather than representational fragility. These findings are supported by graph-theoretic metrics and gene ontology enrichment analyses. Together, our results refine the hourglass framework by localizing developmental constraint to the architecture of gene regulatory networks and demonstrate that language models can extract interpretable biological structure from high-dimensional single-cell data Significance statementUnderstanding how cells coordinate to build complex organisms remains a central challenge in biology. Development is genetically encoded yet context dependent, shaped by interactions among genes, cells, and time. Here, we use a transformer-based language model, Zebraformer, trained on single-cell gene expression data from zebrafish embryos to investigate how regulatory organization evolves. The model captures key features of organismal formation, including transcriptional divergence, anatomical specificity, and reorganization of gene regulatory structure. Attention-derived networks reveal a transient phase of tightly coordinated gene modules with reduced cross-module connectivity during the phylotypic stage, followed by specialization. These findings refine the hourglass hypothesis by localizing constraint to network-level organization and demonstrate that contextual language models uncover core principles of biological organization directly from data.

systems biology↗

bioRxiv2025

Cell atlases and the Developmental Foundations of the Phenotype

It is widely acknowledged that development shapes phenotypes, yet the extent to which genes with similar expression patterns during development lead to equivalent organismal phenotypes when mutated remains unclear. Here, we propose addressing this issue, which we term the [D]evelopment-to- [P]henotype, or [D]-[P], rule, by leveraging single-cell gene expression atlases and phenotypic ontologies, using Caenorhabditis elegans as a model system. This framework quantifies the proportionality between developmental expression and phenotypic similarities, demonstrating that the relationship holds on average. Genes that strongly fulfill the rule exhibit broad "housekeeping" expression and are associated with systemic phenotypes, whereas weak similarities correspond to specific expression patterns and specialized phenotypes. Deviations from the [D]-[P] rule provide insights into developmental divergence and phenotypic degeneracy, highlighting genes with narrow functional roles but systemic phenotypic impact. Furthermore, genes that closely adhere to the rule exhibit the highest pleiotropic impact on organismal traits. Our analysis also identifies cell types, such as ASK neurons, as key mediators of phenotype-specific gene contributions, exemplified by their association with chemosensory behavior and chemotaxis. These findings validate the [D]-[P] rule and underscore the role of cells as critical mediators of the genotypephenotype map, offering a unified framework to understand the developmental origins of phenotypic complexity.

systems biology↗

bioRxiv2024

Polygenic prediction and gene regulation networks

Exploring the degree to which phenotypic variation, influenced by intrinsic nonlinear biological mechanisms, can be accurately captured using statistical methods is essential for advancing our comprehension of complex biological systems and predicting their functionality. Here, we examine this issue by combining a computational model of gene regulation networks with a linear additive prediction model, akin to polygenic scores utilized in genetic analyses. Inspired by the variational framework of quantitative genetics, we create a population of individual networks possessing identical topology yet showcasing diversity in regulatory strengths. By discerning which regulatory connections determine the prediction of phenotypes, we contextualize our findings within the framework of core and peripheral causal determinants, as proposed by the omnigenic model of complex traits. We establish connections between our results and concepts such as global sensitivity and local stability in dynamical systems, alongside the notion of sloppy parameters in biological models. Furthermore, we explore the implications of our investigation for the broader discourse surrounding the role of epistatic interactions in the prediction of complex phenotypes. Author SummaryThis research delves into how well statistical methods can capture phenotypic variation influenced by nonlinear biological mechanisms. The study combines a computational model of gene regulation networks with a linear additive prediction model, similar to polygenic scores used in genetic analysis. By creating a population of individual networks with identical topology but varying regulatory strengths, the research identifies key regulatory connections that predict phenotypes. The findings are framed within the omnigenic model of complex traits, distinguishing core and peripheral causal determinants. The study also links its results to concepts like global sensitivity and local stability in dynamical systems, as well as sloppy parameters in biological models. Additionally, it examines the implications for understanding the role of epistatic interactions in predicting complex phenotypes. This work enhances our understanding of complex biological systems and their functionality.

systems biology↗

bioRxiv2023

Disentangling protein metabolic costs in human cells and tissues.

Many intrinsic functional characteristics of cells and tissues shape their genome-wide expression patterns. But what other factors might also modulate these patterns are not fully known. Here, we revisit the general model of costs in which the protein products of highly expressed genes should be short and made up of biosynthetically cheap amino acids. We first use single-cell expression data from a collection of human cell types to confirm this model with a twist: the most highly expressed proteins tend to be particularly short and use expensive amino acids. By clustering how these two factors change with expression across all cell types, we identified a set of archetypal profiles that uniquely balance costs and that occur at different proportion across cell types. Similar profiles were also found by examining the expression data of tissues, which allowed us to recognize those following a more or less costly strategy. We then asked how this model might delineate the expression changes seen in a tumor relative to its normal solid tissue, as it has been argued that energy constraints determine cancer progression. We discovered a strong signal for the overexpression of biosynthetically cheap compact genes in cancer tissues. Our work highlights how both aspects of the metabolic cost of a protein, length and amino acid biosynthesis, represent valuable measures for understanding the different levels of biological organization and also the differences between health and disease.

systems biology↗

bioRxiv2022

The limitations of phenotype prediction in metabolism

Phenotype prediction is at the core of many questions in biology. Prediction is frequently attained by determining statistical associations between genetic and phenotypic variation, ignoring the exact processes causing the phenotype. Here, we present a framework based on genome-scale metabolic reconstructions to reveal the mechanisms behind the associations. We compute a polygenic score (PGS) that identifies a set of enzymes as predictors of growth, the phenotype. This set arises from the synergy of the functional mode of metabolism in a particular environment and its evolutionary history, and is transportable to infer the phenotype across a range of environments. We also find that there exists an optimal genetic variation for predictability and demonstrate how the linear PGS can yet explain phenotypes generated by the underlying nonlinear biochemistry. Thus, the explicit model interprets the black-box statistical associations of the genotype-to-phenotype map and helps uncover what limits prediction in metabolism.

systems biology↗

bioRxiv2021

Predicting the fitness costs of complex mutations

The fitness cost of complex pleiotropic mutations is generally difficult to assess. On the one hand, it is necessary to identify which molecular properties are directly altered by the mutation. On the other, this alteration modifies the activity of many genetic targets with uncertain consequences. Here, we examine the possibility of addressing these challenges by identifying unique predictors of these costs. To this aim, we consider mutations in the RNA polymerase (RNAP) in Escherichia coli as a model of complex mutations. Changes in RNAP modify the global program of transcriptional regulation, with many consequences. Among others is the difficulty to decouple the direct effect of the mutation from the response of the whole system to such mutation. A problem that we solve quantitatively with data of a set of constitutive genes, which better read the global program. We provide a statistical framework that incorporates the direct effects and other molecular variables linked to this program as predictors, which leads to the identification that some genes are more suitable predictors than others. Therefore, we not only identified which molecular properties best anticipate costs in fitness, but we also present the paradoxical result that, despite pleiotropy, specific genes serve as better predictors. These results have connotations for the understanding of the architecture of robustness in biological systems.

systems biology↗

bioRxiv2019

Genetic buffering and potentiation in metabolism

Cells adjust their metabolism in response to mutations, but how this reprogramming depends on the genetic context is not well known. Specifically, the absence of individual enzymes can affect reprogramming and thus the impact of mutations in cell growth. Here, we examine this issue with an in silico model of Saccharomyces cerevisiaes metabolism. By quantifying the variability in the growth rate of 10000 different mutant metabolisms that accumulated changes in their reaction fluxes, in the presence, or absence, of a specific enzyme, we distinguish a subset of modifier genes serving as buffers or potentiators of variability. We notice that the most potent modifiers refer to the glycolysis pathway and that, more broadly, they show strong pleiotropy and epistasis. Moreover, the evidence that this subset depends on the specific growing condition strengthens its systemic underpinning, a feature only observed before in a simple model of a gene-regulatory network. Some of these enzymes also modulate the effect that biochemical noise and environmental fluctuations produce in growth. Thus, the reorganization of metabolism triggered by mutations has not only direct physiological implications but also changes the influence that other mutations have on growth. This is a general result with implications in the development of cancer therapies based on metabolic inhibitors.

systems biology↗

bioRxiv2019

Tolerance to NADH/NAD+ imbalance anticipates aging and anti-aging interventions

Redox couples coordinate cellular function, but the consequences of their imbalances are unclear. This is somewhat associated with the limitations of their experimental quantification. Here we circumvent these difficulties by presenting a new approach that characterizes fitness-based tolerance profiles to redox couple imbalances using an in silico representation of metabolism. Focusing on the NADH/NAD+ redox couple in yeast, we demonstrate that reductive disequilibria generate metabolic syndromes comparable to those observed in cancer cells. The tolerance of yeast mutants to redox disequilibrium can also explain 30% of the variability in their experimentally measured chronological lifespan. Moreover, by predicting the significance of some metabolites to help stand imbalances, we correctly identify nutrients underlying mechanisms of pathology, lifespan-protecting molecules or caloric restriction mimetics. Tolerance to redox imbalances becomes thus a valid framework to recognize fundamental properties of the aging phenotype while providing a firm biological rationale to assess anti-aging interventions.

systems biology↗

bioRxiv2018

Large-scale analysis of the global transcriptional regulation of bacterial gene expression

Bacterial gene expression depends on the allocation of limited transcriptional resources provided a particular growth rate and growth condition. Early studies in a few genes suggested this global regulation to generate a unifying hyperbolic expression pattern. Here, we developed a large-scale method that generalizes these experiments to quantify the response to growth of over 700 genes that a priori do not exhibit any specific control. We distinguish a core subset following a promoter-specific hyperbolic response. Within this group, we sort genes with regard to their responsiveness to the global regulatory program to show that those with a particularly sensitive linear response are located near the origin of replication. We then find evidence that this genomic architecture is biologically significant by examining position conservation of E. coli genes in 100 bacteria. The response to the transcriptional resources of the cell results consequently in an additional feature contributing to bacterial genome organization.

systems biology↗

bioRxiv2018

Complex genetic and epigenetic regulation deviates gene expression from a global transcriptional program

Environmental or genetic perturbations lead to gene expression changes. While most analyses of these changes emphasize the presence of qualitative differences on just a few genes, we now know that changes are widespread. This large-scale variation has been linked to the exclusive influence of a global transcriptional program determined by the new physiological state of the cell. However, given the sophistication of eukaryotic regulation, we expect to have a complex architecture of specific control affecting this program. Here, we examine this architecture. Using data of Saccharomyces cerevisiae expression in different nutrient conditions, we first propose a five-sector genome partition, which integrates earlier models of resource allocation, as a framework to examine the deviations from the global control. In this scheme, we recognize invariant genes, whose regulation is dominated by physiology, specific genes, which substantially depart from it, and two additional classes that contain the frequently assumed growth-dependent genes. Whereas the invariant class shows a considerable absence of specific regulation, the rest is enriched by regulation at the level of transcription factors (TFs) and epigenetic modulators. We nevertheless find markedly different strategies in how these classes deviate. On the one hand, there are TFs that act in a unique way between partition constituents, and on the other, the action of chromatin modifiers is significantly diverse. The balance between regulatory strategies ultimately modulates the action of the general transcription machinery and therefore limits the possibility of establishing a unifying program of expression change at a genomic scale.

systems biology↗

bioRxiv2018

High-order interactions dominate the functional landscape of microbial consortia

Understanding the link between community composition and function is a major challenge in microbial ecology, with implications for the management of natural microbiomes and the design of synthetic consortia. For this purpose, it is critical to understand the extent to which community functions and properties can be predicted from species traits and what role is played by complex interactions. Inspired by the study of complex genetic interactions and fitness landscapes, here we have examined how the amylolytic function of combinatorial assemblages of seven starch-degrading soil bacteria depends on the functional contributions from each species and their interactions. Filtering our experimental results through the theory of enzyme kinetics, we show that high-order functional interactions dominate the amylolytic rate of our consortia, even though this function is biochemically simple, redundantly distributed in the community, and additive in the absence of inter-species interactions. As the community grows in size, the contribution of high-order functional interactions grows too, making the community function increasingly unpredictable. We can explain the prevalence of high order effects and their sign from the redundancy of ecological interactions in the network, in particular from redundant facilitation towards a high-performing community member. Our results suggest that even simple functions can be dominated by complex interactions, posing challenges for the predictability and bottom-up engineering of ecosystem function in complex multi-species communities.

ecology↗