Biology subjects

Quon, G.

Publications and source records attributed to Quon, G..

15 recordsLinked to original sources

bioRxiv2025

Predicting emergent phenotypes from single cell populations using CELLECTION

Biological systems exhibit emergent phenotypes that arise from the collective behavior of individual components, such as whole-organ functions that arise from the coordinated activity of its individual cells, or organism-level phenotypes that result from the functional interplay of collections of genes in the genome. We present CELLECTION, a deep learning framework that learns to associate subgroups of instances with different emergent phenotypes. We show CELLECTION enables interpretable predictions for heterogeneous tasks, including disease classification, identification of disease-associated cell subtypes, alignment of developmental stages between human model systems, and even predicting relative hand-wing indices across the avian lineage. CELLECTION therefore provides a scalable and flexible framework for identifying key cellular or genetic signatures underlying complex traits in development, disease, and evolution.

genomics↗

bioRxiv2025

A conditional generative model to disentangle morphological variation from batch effects in model organism imaging studies

Identifying organism-level phenotypic variation that arises from genetic variation is a longstanding problem in genetics. Model organisms such as zebrafish are frequently used for genotype to phenotype studies as they can be genetically manipulated, grown and phenotyped for morphological changes in a high throughput manner through automated systems and imaging. However, individuals from model organisms are typically grown in groups: clutches for zebrafish and frogs, and litter for rodents. These clutches act as strong confounders during image analysis, as individuals from different genotypes but grown in the same clutch tend to look more similar to each other than to individuals with the same genotype in other clutches. Existing approaches such as conditional image classification models and domain adaptation approaches perform poorly on addressing these technical batch effects. Here, we propose a conditional latent diffusion model (cLDM) that disentangles these technical batch effects from morphological features by explicitly conditioning on batch-specific variables during generation, enabling targeted separation of technical artifacts from biologically relevant data. This approach enables accurate classification of genotypes using morphological images of individual zebrafish from distinct mutant classes. Furthermore, this approach allows us to efficiently correct for batch effect and demonstrates the versatility of cLDM in tackling domain-specific problems. This work highlights the potential of cLDM to overcome batch effects and extract meaningful features.

bioinformatics↗

bioRxiv2025

sciLaMA: A Single-Cell Representation Learning Framework to Leverage Prior Knowledge from Large Language Models

Single-cell RNA sequencing (scRNA-seq) enables high-resolution exploration of cellular diversity and gene regulation, yet analyzing such data remains challenging due to technical and methodological limitations. Existing task-specific deep generative models like Variational Auto-Encoder (VAE) and its variants struggle to incorporate external biological knowledge, while transformer-based foundational large Language Models (LLMs or large LaMs) face limitations in computational cost and applicability to tabular gene expression data. Here, we introduce sciL-aMA (single-cell interpretable Language Model Adapter), a novel representation learning frame-work that bridges these gaps by integrating static gene embeddings from multimodal LLMs with scRNA-seq tabular data through a paired-VAE architecture. Our approach generates context-aware representations for both cells and genes and outperforms state-of-the-art methods in key single-cell downstream tasks, including batch effect correction, cell clustering, and cell-state-specific gene marker and module identification, while maintaining computational efficiency. sciL-aMA offers a computationally efficient, unified framework for comprehensive single-cell data analysis and biologically interpretable gene module discovery. Source code is available at https://github.com/microsoft/sciLaMA

bioinformatics↗

bioRxiv2024

Gene expansions contributing to human brain evolution

Duplicated genes expanded in the human lineage likely contributed to brain evolution, yet challenges exist in their discovery due to sequence-assembly errors. We used a complete telomere-to-telomere genome sequence to identify 213 human-specific gene families. From these, 362 paralogs were found in all modern human genomes tested and brain transcriptomes, making them top candidates contributing to human-universal brain features. Choosing a subset of paralogs, long-read DNA sequencing of hundreds of modern humans revealed previously hidden signatures of selection, including for T-cell marker CD8B. To understand roles in brain development, we generated zebrafish CRISPR "knockout" models of nine orthologs and introduced mRNA-encoding paralogs, effectively "humanizing" larvae. Our findings implicate two genes in possibly contributing to hallmark features of the human brain: GPR89B in dosage-mediated brain expansion and FRMPD2B in altered synapse signaling. Our holistic approach provides insights and a comprehensive resource for studying gene expansion drivers of human brain evolution.

genomics↗

bioRxiv2024

Sex-specific single cell-level transcriptomic signatures of Rett syndrome disease progression

Dominant X-linked diseases are uncommon due to female X chromosome inactivation (XCI). While random XCI usually protects females against X-linked mutations, Rett syndrome (RTT) is a female neurodevelopmental disorder caused by heterozygous MECP2 mutation. After 6-18 months of typical neurodevelopment, RTT girls undergo poorly understood regression. We performed longitudinal snRNA-seq on cerebral cortex in a construct-relevant Mecp2e1 mutant mouse model of RTT, revealing transcriptional effects of cell type, mosaicism, and sex on progressive disease phenotypes. Across cell types, we observed sex differences in the number of differentially expressed genes (DEGs) with 6x more DEGs in mutant females than males. Unlike males, female DEGs emerged prior to symptoms, were enriched for homeostatic gene pathways in distinct cell types over time, and correlated with disease phenotypes and human RTT cortical cell transcriptomes. Non-cell-autonomous effects were prominent and dynamic across disease progression of Mecp2e1 mutant females, indicating wild-type-expressing cells normalizing transcriptional homeostasis. These results improve understanding of RTT progression and treatment.

genetics↗

bioRxiv2024

Deciphering the History of ERK Activity from Fixed-Cell Immunofluorescence Measurements

The Ras/ERK pathway drives cell proliferation and other oncogenic behaviors, and quantifying its activity in situ is of high interest in cancer diagnosis and therapy. Pathway activation is often assayed by measuring phosphorylated ERK. However, this form of measurement overlooks dynamic aspects of signaling that can only be observed over time. In this study, we combine a live, single-cell ERK biosensor approach with multiplexed immunofluorescence staining of downstream target proteins to ask how well immunostaining captures the dynamic history of ERK activity. Combining linear regression, machine learning, and differential equation models, we develop an interpretive framework for immunostains, in which Fra-1 and pRb levels imply long term activation of ERK signaling, while Egr-1 and c-Myc indicate recent activation. We show that this framework can distinguish different classes of ERK dynamics within a heterogeneous population, providing a tool for annotating ERK dynamics within fixed tissues.

molecular biology↗

bioRxiv2022

Cell adhesion molecules play subclass-specific roles in electrophysiological response and Schizophrenia risk

Multimodal assays such as Patch-seq that simultaneously profile molecular and cellular phenotypes of cells enable the identification of molecular underpinnings of electrophysiological response patterns in neurons. Here we analyzed Patch-seq measurements of thousands of mouse interneurons to identify subclass-specific genes associated with different electrophysiological features. We found extensive subclass specificity: even for the same ephys feature, largely unique sets of genes are associated with that feature in different subclasses. Well established ephys genes such as Reln demonstrated subclass specificity that was previously not reported. Surprisingly, we found that ion channels explained significantly less variation in ephys response across interneurons compared to other genes; in particular, gene sets enriched in cell adhesion genes were amongst the most associated. We found our gene sets associated with action potential dV/dt measurements explained significant heritability of Schizophrenia risk, suggesting a novel role of single neuron electrophysiology in Schizophrenia risk. Finally, we observed significant ephys function switching of cell adhesion molecules across subclasses; the same adhesion molecule was observed to associate with different functional ephys measurements in distinct subclasses and co-express with different genes, suggesting re-purposing of adhesion molecules in different subclasses. Overall, our results yield novel insight into the specificity of roles that individual genes and adhesion molecules play in both single neuron ephys response and Schizophrenia risk.

genomics↗

bioRxiv2022

Alignment-free microbiome-based classification of fresh produce safety and quality

Small samples sizes and loss of up to 50-70% of sequencing data during the data denoising step of preprocessing can limit the statistical power of fresh produce microbiome analyses and prevent detection of important bacterial species associated with produce contamination or quality reduction. Here, we explored an alignment-free analysis strategy using k-mer hashes to identify DNA signatures predictive of produce safety and produce quality, and compared it against the amplicon sequence variant (ASV) strategy that uses a typical denoising step. Random forests (RF)-based classifiers for fresh produce safety and quality using 7-mer hash datasets had significantly higher classification accuracy than those using the ASV datasets. We also demonstrated that the proposed combination of integrating multiple datasets and leveraging an alignment-free 7-mer hash strategy leads to better classification performance for fresh produce safety and quality. Results generated from this study lay the foundation for future studies that wish and need to incorporate and/or compare different microbiome sequencing datasets for the application of machine learning in the area of microbial safety and quality of food.

microbiology↗

bioRxiv2022

Projecting clumped transcriptomes onto single cell atlases to achieve single cell resolution

Multi-modal single cell RNA assays capture RNA content as well as other data modalities, such as spatial cell position or the electrophysiological properties of cells. Compared to dedicated scRNA-seq assays however, they may unintentionally capture RNA from multiple adjacent cells, exhibit lower RNA sequencing depth compared to scRNA-seq, or lack genome-wide RNA measurements. We present scProjection, a method for mapping individual multi-modal RNA measurements to deeply sequenced scRNA-seq atlases to extract cell type-specific, single cell gene expression profiles. We demonstrate several use cases of scProjection, including the identification of spatial motifs from spatial transcriptome assays, distinguishing RNA contributions from neighboring cells in both spatial and multi-modal single cell assays, and imputing expression measurements of un-measured genes from gene markers. scProjection therefore combines the advantages of both multi-modal and scRNA-seq assays to yield precise multi-modal measurements of single cells.

bioinformatics↗

bioRxiv2021

Single-cell RNA-seq analysis reveals lung epithelial cell-specific contributions of Tet1 to allergic inflammation

Tet1 protects against house dust mite (HDM)-induced lung inflammation in mice and alters the lung methylome and transcriptome. In order to explore the role of Tet1 in individual lung epithelial cell types in HDM-induced inflammation, we established a model of HDM-induced lung inflammation in Tet1 knockout and littermate wildtype mice and studied EpCAM+ lung epithelial cells using single-cell RNA-seq analysis. We identified eight EpCAM+ lung epithelial cell types, among which AT2 cells were the most abundant. HDM challenge increased the percentage of alveolar progenitor cells (AP), broncho alveolar stem cells (BAS), and goblet cells, and decreased the percentage of AT2 and ciliated cells. Bulk and cell-type-specific analysis identified genes subject to Tet1 regulation and linked to augmented lung inflammation, including alarms, detoxification enzymes, oxidative stress response genes, and genes in tissue repair. The transcriptomic regulation was accompanied by alterations in TF activities. Trajectory analysis supports that HDM may enhance the differentiation of AP and BAS cells into AT2 cells, independent of Tet1. Collectively, our data showed that lung epithelial cells had common and unique transcriptomic signatures of allergic lung inflammation. Tet1 deletion altered transcriptomic networks in various lung epithelial cells, with an overall effect of promoting allergen-induced lung inflammation.

immunology↗

bioRxiv2021

Interpretable deep generative models for genomics

Deep neural networks implementing generative models for dimensionality reduction have been extensively used for the visualization and analysis of genomic data. One of their key limitations is lack of interpretability: it is challenging to quantitatively identify which input features are used to construct the embedding dimensions, thus preventing insight into why cells are organized in a particular data visualization, for example. Here we present a scalable, interpretable variational autoencoder (siVAE) that is interpretable by design: it learns feature embeddings that guide the interpretation of the cell embeddings in a manner analogous to factor loadings of factor analysis. siVAE is as powerful and nearly as fast to train as the standard VAE but achieves full interpretability of the embedding dimensions. Using siVAE, we exploit a number of connections between dimensionality reduction and gene network inference to identify gene neighborhoods and gene hubs, without the explicit need for gene network inference. We observe a systematic difference in the gene neighborhoods identified by dimensionality reduction methods and gene network inference algorithms in general, suggesting they provide complementary information about the underlying structure of the gene co-expression network. Finally, we apply siVAE to implicitly learn gene networks for individual iPSC lines and uncover a correlation between neuronal differentiation efficiency and loss of co-expression of several mitochondrial complexes, including NADH dehydrogenase, cytochrome C oxidase, and cytochrome b.

genomics↗

bioRxiv2018

scAlign: a tool for alignment, integration and rare cell identification from scRNA-seq data

scRNA-seq dataset integration occurs in different contexts, such as the identification of cell type-specific differences in gene expression across conditions or species, or batch effect correction. We present scAlign, an unsupervised deep learning method for data integration that can incorporate partial, overlapping or a complete set of cell labels, and estimate per-cell differences in gene expression across datasets. scAlign performance is state-of-the-art and robust to cross-dataset variation in cell type-specific expression and cell type composition. We demonstrate that scAlign identifies a rare cell population likely to drive malaria transmission. Our framework is widely applicable to integration challenges in other domains.

bioinformatics↗

bioRxiv2018

Predicting trait regulators by identifying co-localization of DNA binding and GWAS variants in regulatory regions

Genomic regions associated with complex traits and diseases are primarily located in non-coding regions of the genome and have unknown mechanism of action. A critical step to understanding the genetics of complex traits is to fine-map each associated locus; that is, to find the causal variant(s) that underlie genetic associations with a trait. Fine-mapping approaches are currently focused on identifying genomic annotations, such as transcription factor binding sites, which are enriched in direct overlap with candidate causal variants. We introduce CONVERGE, the first computational tool to search for co-localization of GWAS causal variants with transcription factor binding sites in the same regulatory regions, without requiring direct overlap. As a proof of principle, we demonstrate that CONVERGE is able to identify five novel regulators of type 2 diabetes which subsequently validated in knockdown experiments in pancreatic beta cells, while existing fine-mapping methods were unable to find any statistically significant regulators. CONVERGE also recovers more established regulators for total cholesterol compared to other fine-mapping methods. CONVERGE is therefore unique and complementary to existing fine-mapping methods and is useful for exploring the regulatory architecture of complex traits.

bioinformatics↗

bioRxiv2018

Systems-level properties of EGFR/Ras/ERK signalling amplify local signals to generate gene expressionplasticity

The EGFR/Ras/ERK signalling pathway is a driver of cancer cell proliferation and metastasis in tumours that exhibit high cell-to-cell heterogeneity. While the signalling activity of this pathway is frequently amplified in tumours, it is not understood how the kinetic aspects of its activation in tumours differ from normal cellular signalling. Using live-cell reporters of ERK signalling in the breast cancer progression series HMT-3522, we found that ERK activity in invasive cells is similar in amplitude to isogenic non-malignant cells but is highly dynamic and more disordered, leading to more heterogeneous expression of ERK target genes. Our analysis reveals that this diversification arises from systems-level functions of the pathway, including intracellular amplification of amphiregulin-mediated paracrine signalling and differential kinetic filtering by genes including Fra-1, c-Myc, and Egr1. Our findings establish a mechanism for the generation of non-genetic tumour cell plasticity arising from the specific quantitative properties of a signal transduction pathway.

cancer biology↗

bioRxiv2018

Gene detection models outperform gene expression for large-scale scRNA-seq analysis

Technical variation in feature measurements such as gene expression and locus accessibility is a key challenge of large-scale single cell genomic datasets. We show that this technical variation in both scRNA-seq and scATAC-seq datasets can be mitigated by performing analysis on feature detection patterns alone and ignoring feature quantification measurements. This result holds when datasets have low detection noise relative to quantification noise. We demonstrate state-of-the-art performance of detection pattern models using our new framework, scBFA, for both cell type identification and trajectory inference. Performance gains can also be realized in one line of R code in existing pipelines.

bioinformatics↗