Biology subjects

Ursu, E.

Publications and source records attributed to Ursu, E..

4 recordsLinked to original sources

bioRxiv2025

Morphogen and juxtacrine signalling dynamically integrate to specify cell fates with single-cell resolution

Morphogen gradients guide tissue patterning but do not act in isolation. How they integrate with other signalling modalities, like juxtacrine signalling, and how these integrations influence pattern resolution and robustness remain unclear. We address this in the Drosophila lamina, where columns of precursors are patterned with single-cell resolution into motion-processing neurons, dependent on a photoreceptor-derived Hedgehog gradient and glial-orchestrated differentiation. Combining experiments and theory, we show that glial-induced ERK activity drives Delta expression in lamina precursors, generating a graded Notch activity pattern. Notch restricts Hedgehog morphogen relay and enhances positional information. Glia act as timekeepers, scheduling ERK-driven differentiation after Hedgehog and Notch patterns are established. Thus, Notch and ERK dynamically integrate with Hedgehog to encode positional information, enabling reproducible cell fate patterning with single-cell resolution.

developmental biology↗

bioRxiv2024

Training data composition determines machine learning generalization and biological rule discovery

Supervised machine learning models depend on training datasets with positive and negative examples. Therefore, dataset composition directly impacts model performance and bias. Given the importance of machine learning for immunotherapeutic design, we examined how different negative class definitions affect model generalization and rule discovery for antibody-antigen binding. Using synthetic structure-based binding data, we evaluated models trained with various definitions of negative sets. Our findings reveal that high out-of-distribution performance can be achieved when the negative dataset contains more similar samples to the positive dataset despite a lower within-distribution performance. Furthermore, leveraging ground truth information, we show that binding rules discovered as associated with positive data change based on the negative data used. Validation on experimental data supported simulation-based observations. This work underscores the role of dataset composition, including negative data selection, in creating robust, generalizable, and biology-aware sequence-based ML models.

bioinformatics↗

bioRxiv2022

Drug prediction for reversing AD/PD transcriptional profiles using an aging of systems-centric approach

Age-related pathologies are so widely presented in old age that in most cases they are hardly distinguishable at the molecular level from the so-called normal aging. Both aging and age-related diseases are characterized by a wide range of transcriptional and epigenetic changes that underlie the physiological or pathological phenotype, with plenty of overlap in their signatures, but also with differences. In most pathological conditions it is rather the dysregulation of a complex network of genes than a problem with a single gene dysregulation that causes its emergence or progression, and aging differently gives a "predisposition" towards an age-related pathology or another, or in a favorable situation towards none. The important question is how similar are the transcriptional changes during "healthy" aging with those that occur in age-related diseases. In this study, we explore gene expression data to answer this question and aim to predict which drugs and compounds could have a reversing effect on their common drift.

bioinformatics↗

bioRxiv2021

scAgeCom: a murine atlas of age-related changes in intercellular communication inferred with the package scDiffCom

Dysregulation of intercellular communication is a well-established hallmark of aging. To better understand how this process contributes to the aging phenotype, we built scAgeCom, a comprehensive atlas presenting how cell-type to cell-type interactions vary with age in 23 mouse tissues. We first created an R package, scDiffCom, designed to perform differential intercellular communication analysis between two conditions of interest in any mouse or human single-cell RNA-seq dataset. The package relies on its own list of curated ligand-receptor interactions compiled from seven established studies. We applied this tool to single-cell transcriptomics data from the Tabula Muris Senis consortium and the Calico murine aging cell atlas. All the results can be accessed online, using a user-friendly, interactive web application (https://scagecom.org). The most widespread changes we observed include upregulation of immune system processes, inflammation and lipid metabolism, and downregulation of extracellular matrix organization, growth, development and angiogenesis. More specific interpretations are also provided.

bioinformatics↗