Biology subjects

Caicedo, J.

Publications and source records attributed to Caicedo, J..

5 recordsLinked to original sources

bioRxiv2025

Image-Based Profiling of Induced Trophoblast Stem Cells Identifies Signatures Associated with Sex, Schizophrenia Genomic Risk and Placental Stress

Schizophrenia (SCZ) is a neurodevelopmental disorder where both genetic and environmental risks converge during pregnancy. Recent studies have highlighted the importance of placental biology in influencing risk of developing SCZ. However, the pathways by which genetic risk factors for SCZ interact with environmental influences to alter placental development are poorly understood. Through image-based profiling using Cell Painting, we leveraged trophoblast cultures derived from human induced pluripotent stem cells (hiPSCs) from male and female SCZ and neurotypical donors with varying placental genomic risk scores (PlacGRS) to explore the developmental dynamics of placental cells under normal growth and hypoxic stress. We employed both classical (e.g., CellProfiler) and deep learning feature extraction combined with downstream supervised machine learning to analyze high-dimensional data obtained from this hiPSC-derived model system, highlighting that these approaches overcome the inherent line-to-line variability in phenotypic analysis. Our findings reveal a salient nucleus-localized SCZ risk signature across cell lines, along with clear sexual dimorphism. This research underscores the capability of hiPSC-derived placenta models to elucidate complex interactions between genetic risk and environmental factors implicated in the neurodevelopment of SCZ, paving the way for future studies aimed at developing targeted therapeutic and prevention strategies.

neuroscience↗

bioRxiv2025

Portable transcranial therapeutic ultrasound enhances targeted gene delivery for Parkinson's disease: from rodent models to non-human primates

Gene therapy for neurodegenerative diseases faces significant challenges due to the blood-brain barrier (BBB), which limits drug delivery to the central nervous system (CNS). While clinical trials for Parkinsons disease (PD) have progressed, administration of vectors expressing enzymatic or neurotrophic factor transgenes have required extensive optimization of the delivery method to achieve potentially therapeutic levels of transgene expression. Focused ultrasound (FUS) combined with microbubbles has emerged as a promising non-invasive strategy to transiently open the BBB for targeted gene delivery via viral nanocarriers including recombinant adeno-associated viruses (AAVs). However, key factors influencing FUS-mediated AAV delivery, including dose distribution and therapeutic efficacy, remain underexplored in non-human primates (NHPs). Here, we evaluated the feasibility of AAV9-CAG-GFP delivery using two portable therapeutic ultrasound modalities: ultrasound-guided, spherically-focused FUS (USgFUS) and a novel low-frequency linear array configuration for imaging and therapy called theranostic ultrasound (ThUS). In mice, FUS-sonicated regions exhibited a 25-fold increase in AAV9 biodistribution compared to systemic injection alone. Extending this approach to NHPs, we observed up to a 200-fold increase in AAV9 DNA in treated brain regions, including PD-relevant structures. In assessing the translational therapeutic potential of this technique, ThUS-mediated AAV9-hSyn-hNTRN (human neurturin) delivery in a toxin mouse model of PD facilitated the rescue of up to 80% and 75% of degenerated dopaminergic neurons in the substantia nigra and striatum, respectively. These findings demonstrate that portable ultrasound technologies can non-invasively enhance AAV9 delivery to targeted brain regions in both mice and NHPs relative to what can be achieved with intravenous (IV) delivery of the same capsid alone. With further development, these approaches may offer a clinically viable, non-invasive alternative for gene therapy in neurodegenerative diseases. One sentence summaryBBB opening with portable therapeutic ultrasound non-invasively increased viral gene delivery to the brain after systemic AAV vector administration in mice and rhesus macaques.

bioengineering↗

bioRxiv2022

Deciphering causal genomic templates of complex molecular phenotypes

The genetic code is a formal principle that determines which proteins an organism can produce from only its genome sequence, without mechanistic modeling. Whether similar formal principles govern the relationship between genome sequence and phenotype across scales - from molecules to cells to tissues - is unknown. Here, we show that a single formal principle - structural correspondence -- underlies the relationship between phenotype and genome sequence across scales. We represent phenotypes and the genome as graphs and find mappings between them using structure preservation as the sole constraint. Combinatorial richness in phenotypes more tightly constrains which mappings preserve that structure. Thus, phenotypic structure predicts genetic associations independently of covariation with genotype. This principle rediscovers the amino acid code without prior knowledge of translation or coding sequences, using just one protein and genome sequence as input. We benchmark this principle: applied to phenotypes at the cell, tissue and organ scales, the mappings correctly predict established associations and are driven by transcription factor motifs. Applied to cancer tissue images, we find regulators of spatial gene expression in immune cells. We thus offer a first-principles framework to relate genome sequence with phenotypic structure and guide mechanistic discovery across scales.

genetics↗

bioRxiv2021

Morphology and gene expression profiling provide complementary information for mapping cell state

Morphological and gene expression profiling can cost-effectively capture thousands of features in thousands of samples across perturbations by disease, mutation, or drug treatments, but it is unclear to what extent the two modalities capture overlapping versus complementary information. Here, using both the L1000 and Cell Painting assays to profile gene expression and cell morphology, respectively, we perturb A549 lung cancer cells with 1,327 small molecules from the Drug Repurposing Hub across six doses, providing a data resource including dose-response data from both assays. The two assays capture both shared and complementary information for mapping cell state. Cell Painting profiles from compound perturbations are more reproducible and show more diversity, but measure fewer distinct groups of features. Applying unsupervised and supervised methods to predict compound mechanisms of action (MOA) and gene targets, we find that the two assays provide a partially shared, but also a complementary view of drug mechanisms. Given the numerous applications of profiling in biology, our analyses provide guidance for planning experiments that profile cells for detecting distinct cell types, disease phenotypes, and response to chemical or genetic perturbations.

bioinformatics↗

bioRxiv2018

Evaluation of Deep Learning Strategies for Nucleus Segmentation in Fluorescence Images

Identifying nuclei is often a critical first step in analyzing microscopy images of cells, and classical image processing algorithms are most commonly used for this task. Recent developments in deep learning can yield superior accuracy, but typical evaluation metrics for nucleus segmentation do not satisfactorily capture error modes that are relevant in cellular images. Besides, large image data sets with ground truth for evaluation have been limiting. We present an evaluation framework to measure accuracy, types of errors, and computational efficiency; and use it to compare two deep learning strategies (U-Net and DeepCell) alongside a classical approach implemented in CellProfiler. We publicly release a set of 23,165 manually annotated nuclei and source code to reproduce experiments. Our results show that U-Net outperforms both pixel-wise classification networks and classical algorithms. Also, our evaluation framework shows that deep learning improves accuracy and reduces the number of biologically relevant errors by half.

bioinformatics↗