Biology subjects

Baghdassarian, H.

Publications and source records attributed to Baghdassarian, H..

6 recordsLinked to original sources

bioRxiv2025

Biologically informed neural network models are robust to spurious interactions via self-pruning

Computational models of cellular networks hold promise to uncover disease mechanisms and guide therapeutic strategies. Biology-informed neural networks (BINNs) is an emerging approach to create such models by combining the predictive power of deep learning with prior knowledge, a vital aspect of biological research. The architectures of BINNs enforces a network structure from which mechanism can ideally be inferred. However, a key challenge is to evaluate the reliability of these models, as cells are inherently complex, involving intricate and sometimes unknown interactions. Currently, analysis mainly focuses on selected pathways rather than a more comprehensive perspective. In this work we demonstrate an alternative holistic approach: we measure to which extent purposefully introduced spurious interactions are down-weighted by a BINN during training (self-pruning). The metric suggested RRA (Relative Residual Area) allows for direct distribution comparison with perfect self-pruning achieved at zero and a failure to self-prune if above one. To enable rapid testing, we updated LEMBAS (Large-scale knowledge-EMBedded Artificial Signaling-networks), our recurrent neural network framework for intracellular signaling dynamics, with full GPU acceleration. Our implementation achieves a >7-fold speedup compared to the original while preserving predictive accuracy. We evaluated self-pruning in 3 different datasets and found that when spurious interactions are introduced at random, the model prunes these to a larger extent than those from the prior knowledge network (PKN), provided the model is regularized with a sufficiently large L2 norm. This suggests that BINNs can be robust to uncertainty in the PKN. Implementation and applicationOur implementation of LEMBAS is freely available under a MIT license at https://github.com/AvlantNilssonLab/LEMBAS_GPU. The models and results to generate the figures can be downloaded through https://zenodo.org/records/17425598.

systems biology↗

bioRxiv2024

A reconstruction of the mammalian secretory pathway identifies mechanisms regulating antibody production

The secretory pathway processes >30% of mammalian proteins, orchestrating their synthesis, modification, trafficking, and quality control. However, its complexity-- spanning multiple organelles and dependent on coordinated protein interactions--limits our ability to decipher how protein secretion is controlled in biomedical and biotechnological applications. To advance such research, we present secRecon--a comprehensive reconstruction of the mammalian secretory pathway, comprising 1,127 manually curated genes organized within an ontology of 77 secretory process terms, annotated with functional roles, subcellular localization, protein interactions, and complex composition. Using secRecon to integrate multi-omics data, we identified distinct secretory topologies in antibody-producing plasma cells compared to CHO cells. Genes within proteostasis, translocation, and N-glycosylation are deficient in CHO cells, highlighting them as potential engineering targets to boost secretion capacity. Applying secRecon to single-cell transcriptomics and SEC-seq data, we uncovered secretory pathway signatures underlying secretion diversity among IgG-secreting plasma cells. Different transcriptomic clusters had unique secretory phenotypes characterized by variations in the unfolded protein response (UPR), endoplasmic reticulum-associated degradation (ERAD), and vesicle trafficking pathways. Additionally, we discovered specific secretory machinery genes as new markers for plasma cell differentiation. These findings demonstrate secRecon can identify mechanisms regulating protein secretion and guide diverse studies in biomedical research and biotechnology. Graphical Abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=73 SRC="FIGDIR/small/623668v1_ufig1.gif" ALT="Figure 1"> View larger version (24K): org.highwire.dtl.DTLVardef@f9f865org.highwire.dtl.DTLVardef@122db7forg.highwire.dtl.DTLVardef@1788a1aorg.highwire.dtl.DTLVardef@b2b2bd_HPS_FORMAT_FIGEXP M_FIG C_FIG

systems biology↗

bioRxiv2022

Unraveling the coordinated dynamics of protein- and metabolite-mediated cell-cell communication

SummaryCell-cell communication dynamically changes across time while involving diverse cell populations and ligand types such as proteins and metabolites. While single-cell transcriptomics enables its inference, existing tools typically analyze ligand types separately and overlook their coordinated activity. Here, we present Tensor-cell2cell v2, a computational tool that can jointly analyze protein- and metabolite-mediated communication over time using coupled tensor component analysis, while preserving each modality of inferred communication scores independently, as well as their data structures and distributions. Applied to brain organoid development, Tensor-cell2cell v2 uncovers dynamic, coordinated communication programs involving key proteins and metabolites across relevant cell types across specific time points. Availability and implementationTensor-cell2cell v2 and its new coupled tensor component analysis are implemented in Python and available as part of the cell2cell framework at https://github.com/earmingol/cell2cell. This python library is available on PyPI. Analyses of this manuscript can be reproduced in a Code Ocean capsule at https://doi.org/10.24433/CO.0061424.v1 and online tutorials can be found at https://cell2cell.readthedocs.io. Supplementary informationSupplementary data are available at bioRxiv online.

bioinformatics↗

bioRxiv2021

Context-aware deconvolution of cell-cell communication with Tensor-cell2cell

Cell interactions determine phenotypes, and intercellular communication is shaped by cellular contexts such as disease state, organismal life stage, and tissue microenvironment. Single-cell technologies measure the molecules mediating cell-cell communication, and emerging computational tools can exploit these data to decipher intercellular communication. However, current methods either disregard cellular context or rely on simple pairwise comparisons between samples, thus limiting the ability to decipher complex cell-cell communication across multiple time points, levels of disease severity, or spatial contexts. Here we present Tensor-cell2cell, an unsupervised method using tensor decomposition, which is the first strategy to decipher context-driven intercellular communication by simultaneously accounting for multiple stages, states, or locations of the cells. To do so, Tensor-cell2cell uncovers context-driven patterns of communication associated with different phenotypic states and determined by unique combinations of cell types and ligand-receptor pairs. As such, Tensor-cell2cell robustly improves upon and extends the analytical capabilities of existing tools. We show Tensor-cell2cell can identify multiple modules associated with distinct communication processes (e.g., participating cell-cell and ligand receptor pairs) linked to COVID-19 severities and Autism Spectrum Disorder. Thus, we introduce an effective and easy-to-use strategy for understanding complex communication patterns across diverse conditions.

bioinformatics↗

bioRxiv2020

Inferring the spatial code of cell-cell interactions and communication across a whole animal body

Cell-cell interactions shape cellular function and ultimately organismal phenotype. However, the code embedded in the molecular interactions driving and sustaining the spatial organization of cells remains to be elucidated. Here we present a computational framework to infer the spatial code underlying cell-cell interactions from the transcriptomes of the cell types across the whole body of a multicellular organism. As core of this framework, we introduce our tool cell2cell, which uses the coexpression of ligand-receptor pairs to compute the potential for intercellular interactions, and we test it across the Caenorhabditis elegans body. Leveraging a 3D atlas of C. elegans cells, we also implement a genetic algorithm to identify the ligand-receptor pairs most informative of the spatial organization of cells. Validating the spatial code extracted with this strategy, the resulting intercellular distances are negatively correlated with the inferred cell-cell interactions. Furthermore, for selected cell-cell and ligand-receptor pairs, we experimentally confirm the communicatory behavior inferred with cell2cell and the genetic algorithm. Thus, our framework helps identify a code that predicts the spatial organization of cells across a whole-animal body.

bioinformatics↗

bioRxiv2020

Learned adaptive multiphoton illumination microscopy

Multiphoton microscopy is a powerful technique for deep in vivo imaging in scattering samples. However, it requires precise, sample-dependent increases in excitation power with depth in order to maintain signal while minimizing photodamage. We show that cells with identical fluorescent labels imaged in situ can be used to train a physics-based machine learning model that solves this problem. After this training has been performed, the correct illumination power can be predicted and adaptively adjusted at each point in a 3D volume on subsequent samples as a function of the samples shape, without the need for specialized fluorescent labelling. We use this technique for in vivo imaging of immune responses in mouse lymph nodes following vaccination, with imaging volumes 2-3 orders of magnitude larger than previously reported. We achieve visualization of physiologically realistic numbers of antigen-specific T cells for the first time, and demonstrate changes in the global organization and motility of dendritic cell networks during the early stages of the immune response.

immunology↗