Biology subjects

Engelhardt, B. E.

Publications and source records attributed to Engelhardt, B. E..

At least 19 recordsLinked to original sources

bioRxiv2026

Scalable multi-group nonnegative spatial factorization for spatial genomics data with cell-type heterogeneity

Spatial transcriptomics (ST) technologies enable the study of gene expression within the spatial context of tissues, providing insights into tissue structure, cellular interactions, and disease progression. However, existing dimension reduction methods often overlook spatial information or struggle to distinguish spatial gene patterns from those driven by cell-type differences, limiting biological interpretability by convolving differences in gene expression patterns with differences in cell-type proportions. To address these challenges, we introduce the scalable multi-group nonnegative spatial factorization (smNSF), a computationally-tractable probabilistic framework that integrates spatial coordinates and cell-type labels into a unified matrix factorization model. By using multi-group Gaussian processes (MGGPs) as priors, our model captures complex spatial variation in a cell-type specific way while enforcing nonnegativity to enhance interpretability. We develop a variational inference framework for MGGPs that supports scalable optimization and improves the numerical stability of smNSF. Across seven spatial transcriptomics datasets spanning diverse technologies and tissues, smNSF recovers sparse, interpretable spatial factors and, through its cell-type conditional posteriors, organizes them into cell-type enriched, cell-type specific, and universal spatial programs that are not apparent from marginal factors alone. Given cell-type labels in ST data, smNSF enables cell-type aware spatial decompositions and supports cell-type conditional posteriors for in silico exploration of relationships between spatial patterns and cellular identity. Author summaryMost current analysis methods for spatial transcriptomics either ignore spatial structure or fail to separate gene-driven spatial patterns from cell-type driven differences. In this work, we develop a method that uses spatial coordinates and cell-type labels together to better uncover patterns of gene expression. Our approach, scalable multi-group nonnegative spatial factorization (smNSF), uses Gaussian processes to model spatial structure, which we extend to capture both spatial structure and cell type within a unified framework called multi-group Gaussian processes. Applying smNSF to spatial transcriptomics datasets from mouse brain and human tissues, we find that conditioning on different cell types reveals spatial patterns that are invisible in standard analyses: some cell types suppress a pattern, others sharpen it, and some reveal structure that only emerges after conditioning. This helps us understand how cell-type specific spatial programs contribute to tissue organization. To make this analysis tractable on the scale of modern spatial experiments, we also introduce a new computational approximation for Gaussian processes that is directly applicable in principle to other latent variable GP models. Together, these tools help disentangle biological sources of variation and support in silico exploration of how gene expression might change under different tissue or cell-type compositions.

genomics↗

bioRxiv2026

Machine Learning-Guided Discovery of Bacterial-Selective Membrane-Active Compounds Reveals Mechanistic Bias in Antibiotic Training Datasets

The rise of antibiotic resistance necessitates the discovery of antibacterial compounds with novel mechanisms of action (MoAs). Recent machine learning approaches have shown promise in antibacterial compound discovery, but often identify derivatives of known antibiotic classes rather than mechanistically novel compounds. Previous approaches applied Tanimoto similarity filters at the end of screening pipelines, but this method has substantial drawbacks: Tanimoto similarity can be misleading in chemical space, and post-hoc filtering does not influence what activity models learn to prioritize. Here, we present a machine learning pipeline that addresses chemical novelty upfront by employing an XGBoost-based MoA classifier to explicitly prioritize compounds predicted to have mechanisms distinct from known antibiotic classes, combined with graph neural networks for antibacterial activity and toxicity prediction. Applied to the Zinc20 database, our approach successfully identified non-toxic antibacterial compounds structurally distinct from known antibiotics. Notably, the majority of these hits exhibited membrane-targeting activity with selectivity for bacterial cells over mammalian cells, suggesting potential for next-generation membrane-active antibiotics. However, we did not identify compounds with novel protein targets. Systematic analysis revealed that this limitation stems from mechanistic bias in training data rather than model architecture. Specifically, our activity model learned to preferentially score compounds similar to specific groups in the training data, thus overrepresenting certain MoA classes including membrane-active compounds. Even substantial model architecture and training data enhancements did not overcome this constraint. Our findings demonstrate that the primary bottleneck for discovering mechanistically novel antibiotics is the scarcity of diverse, mechanistically-annotated training data. This work provides both a methodological framework for mechanism-aware screening and critical insights into data requirements for genuinely novel antibiotic discovery.

bioinformatics↗

bioRxiv2026

Using image classifiers to predict CMT2A disease-relevant mitochondrial motility phenotypes in iPSC motor neurons

Charcot-Marie-Tooth disease type 2A (CMT2A) is a genetic disease characterized by autosomal dominant MFN2 mutations and dysregulated mitochondrial trafficking. While there is currently no FDA-approved CMT2A therapy, the recent development of iPSC motor neuron model systems, high-throughput imaging platforms, and CRISPR-based gene editing technologies holds promise for screening new therapies at scale in vitro. A critical roadblock for therapeutic screening is the development of scalable and robust computational methods to assess the mitochondrial trafficking phenotypes, healthy or diseased, of each iPSC motor neuron sample. To address this gap, we developed a vision transformer (ViT) based classification framework that predicts disease phenotypes using kymographs, an image transformation that captures particle movement along a prespecified path, such as mitochondrial movement along axons. We show that our classification approach more accurately discriminates healthy MFN2 wild-type (WT) from diseased MFN2 R364W-mutant (R364W) iPSCs than alternative summary statistics, such as mitochondrial speed and fraction of stationary mitochondria that are directly extracted from kymographs. Furthermore, we show that our model maintains high accuracy when deployed on a biological replicate holdout dataset. An analysis of ViT patch embeddings of the kymographs shows that mitochondria with highly variable sizes and many intersection events most strongly associate with R364W diseased kymographs. The computational approach demonstrated in this paper has broad applicability for future high-throughput screens where organelle trafficking along axons plays a key role in disease pathogenesis.

cell biology↗

bioRxiv2026

Differential Methylation by Early Life Adversity in the Future of Families Child Wellbeing Study

Early life adversity (ELA) has a well-established link to mental health disorders later in life, yet the molecular mechanisms behind this relationship are incompletely understood. The Future of Families Child Wellbeing Study (FFCWS) provides an opportunity to examine experience encoding in the genome through functional changes in DNA methylation (DNAm) in a cohort enriched in subjects exposed to ELA. We investigated epigenome-wide differences in DNAm across thirteen early-life exposures in salivary samples from FFCWS participants. We calculated differential methylation associations with disease-linked genetic variants, evaluated tissue-specific gene expression, and assessed the persistence of DNAm changes from ages 9 to 15 years. Using data from the mSTARR-seq assay, we characterized methylation-dependent regulatory activity. Differential methylation in the FFCWS validated prior results and identified new genomic regions associated with child adversity. Differential methylation occurs in genomic regions likely to impact gene expression, and affected genes are expressed in disease-relevant tissues. We also identified association of genetic variants associated with downstream disorders near differential methylation, including depression, alcohol and substance use, and anxiety disorders. Overall, cumulative ELA is associated with specific DNAm changes, functional regulation, and persistence over time. Our findings indicate that ELA-associated differential methylation in the FFCWS does not simply occur at random, but in genomic regions that are functional. Our results support the conclusion that altered DNAm represents a biological link between early life experience and later health outcomes.

genomics↗

bioRxiv2025

k-spaces: Mixtures of Gaussian latent variable models

Principal component analysis (PCA) and k-means clustering are two seemingly different methods for dimension reduction and clustering, respectively, but can be understood as special cases of inference in a Gaussian latent variable model framework. We leverage this insight to develop a probabilistic framework and methods for simultaneous dimension reduction, clustering, and latent space learning that are efficient and interpretable, and that can replace current ad hoc combinations of PCA and clustering. The algorithm, k-spaces, has broad applicability, which we demonstrate in several distinct genomic settings. In particular, we show how k-spaces can be used to model gene expression in quantitative hybridization chain reaction (qHCR) images, for inference in epigenomics, and for dimension reduction of single-cell RNA-sequencing data.

bioinformatics↗

bioRxiv2025

SPARK: in silico simulations for benchmarking nascent RNA sequencing experiments

Nascent RNA sequencing offers profound insights into transcriptional dynamics, yet there are substantial challenges to analyzing these data. The development of proper computational tools necessitates realistic benchmarking datasets that reflect biological variability and technical biases. We present simulated pre-mRNA and RNA kinetics (SPARK), a versatile in silico framework for generating reads across nascent RNA sequencing approaches. SPARK simulates the process of transcription -- allowing for variable elongation rates and pausing events -- and key experimental features. SPARK provides a comprehensive platform for computational development and benchmarking in nascent RNA genomics.

genomics↗

bioRxiv2025

Live-cell analyses with unsegmented images to study cancer cell response to modified T cell therapy

Live-cell imaging (LCI) of modified T cells co-cultured with cancer cells is commonly used to quantify T cell anti-cancer function. Videos captured by LCI show complex multi-cell behavioral phenotypes that go beyond simple cancer cell fluorescence measurements. Here, we develop an unsupervised analysis workflow to characterize LCI data generated using the Incucyte imaging platform. Unlike most LCI analyses, we avoid cell segmentation due to the low spatiotemporal resolution of the LCI videos and high levels of cell-cell contact. Instead, we develop methods that identify global aggregation patterns and local cellular keypoints to characterize the multicellular interactions that determine cancer cell sensitivity to, or escape from, T cell surveillance. We demonstrate our segmentation-free live-cell behavioral analysis (SF-LCBA) methods on TCR T cells from four donors with varying proportions of cells with a beneficial RASA2 knockout and effector-to-target initial concentrations in a co-culture with A375 melanoma cells. We find that different T cell modifications affect the spatiotemporal dynamics of multicellular aggregate formation. In particular, we show that fewer and smaller cancer cell aggregates form at high ratios of effector T cells to target cancer cells and high titrations of T cells with RASA2 knockouts. Our SF-LCBA method identifies, characterizes, and tracks cellular aggregate formation in datasets that are unsuitable for cell segmentation and tracking, opening the door to more therapeutically-relevant measurements of modified T cell therapy cell behavioral phenotypes from LCI data.

cancer biology↗

bioRxiv2025

Neighborhood nonnegative matrix factorization identifies patterns and spatially-variable genes in large-scale spatial transcriptomics data

Tissues consist of multi-cellular neighborhoods in which different cell types express correlated gene programs due to shared signaling environments. Methods for identifying these spatial neighborhoods may be powerful, but currently do not scale to existing data sets of millions of cells and often artificially divide tissues into distinct neighborhoods with hard borders. To better identify multi-cellular microenvironments with shared gene programs in large-scale spatial genomics data, we developed a method that combines nonnegative matrix factorization (NMF) with Gaussian smoothing across cells in space. Our spatially-aware dimension reduction, neighborhood NMF (NNMF), identifies known and unknown interactions among diverse cell types organized into complex patterns, from localized structures to broad tissue regions. NNMF has many advantages over currently available methods, including the ability to run on modern large-scale data with thousands of features and multiple tissue samples and with millions of cells. Furthermore, our method is based on probabilistic NMF, which produces soft clusters of landscape signatures that can be understood as overlapping spatially-organized multicellular gene expression programs, allowing more biologically-complete interpretations than overly-simplistic hard clustering methods. In a benchmark dataset of a diverse set of spatial gene expression data with expert tissue labels, compared against related methods, NNMF with K-nearest neighbors clustering shows excellent performance even on hard clustering tasks. On MERFISH human colorectal cancer data, NNMF identifies several immunologically relevant multicellular interaction networks and scales to these data sets with million of cells. NNMF is implemented as an R package available at https://github.com/ragnhildlaursen/NNMF.

genomics↗

bioRxiv2025

Nearest-neighbor nonnegative spatial factorization to study spatial and temporal transcriptomics

Nonnegative spatial factorization (NSF) is a spatially-aware factorization method that uses Gaussian processes (GPs) as spatial priors in a Poisson latent factor model to robustly identify interpretable, parts-based representations in spatial transcriptomics data. However, NSF scales poorly with modern datasets due to the computational complexity of Gaussian processes, which scales cubically with the number of points used for inference O(N 3). To address this limitation, we propose a modified version of NSF that leverages variational nearest neighbor Gaussian processes (VNNGPs), resulting in a substantial reduction in inference complexity from O(NM 2) in the current version of NSF to O(MNK2) for M inducing points, N total points and K nearest neighbors. Our method, nearest-neighbor NSF (NNNSF), is benchmarked on synthetic and real-world spatial and temporal transcriptomics datasets. Experimental results demonstrate that NNNSF achieves linear scaling with the number of neighbors and points used for inference in contrast with NSF, which has exponential computational complexity as the number of points used in inference increases. By restricting covariance calculations to the K-nearest neighbors of the points used in inference, NNNSF allows the use of more inducing points, leading to lower reconstruction loss. Nearest-neighbor NSF (NNNSF), which replaces standard variational inference with inducing points in NSF with the VNNGP, leads to a computationally efficient and scalable version of NSF that can be applied to large existing and forthcoming spatial genomics data. We added VNNGP and NNNSF to the GPZoo package, an an ongoing open source project developing a modular Gaussian process library in Python making use of the PyTorch interface. Source code and demonstrations are available at https://github.com/luisdiaz1997/GPzoo/tree/main.

bioinformatics↗

bioRxiv2024

Cellular behavior analysis from live-cell imaging of TCR T cell-cancer cell interactions

T cell therapies, such as chimeric antigen receptor (CAR) T cells and T cell receptor (TCR) T cells, are a growing class of anti-cancer treatments. However, expansion to novel indications and beyond last-line treatment requires engineering cells dynamic population behaviors. Here we develop the tools for cellular behavior analysis of T cells from live-cell imaging, a common and inexpensive experimental setup used to evaluate engineered T cells. We first develop a state-of-the-art segmentation and tracking pipeline, Caliban, based on human-in-the-loop deep learning. We then build the Occident pipeline to collect a catalog of phenotypes that characterize cell populations, morphology, movement, and interactions in co-cultures of modified T cells and antigen-presenting tumor cells. We use Caliban and Occident to interrogate how interactions between T cells and cancer cells differ when beneficial knock-outs of RASA2 and CUL5 are introduced into TCR T cells. We apply spatiotemporal models to quantify T cell recruitment and proliferation after interactions with cancer cells. We discover that, compared to a safe harbor knockout control, RASA2 knockout T cells have longer interaction times with cancer cells leading to greater T cell activation and killing efficacy, while CUL5 knockout T cells have increased proliferation rates leading to greater numbers of T cells for hunting. Together, segmentation and tracking from Caliban and phenotype quantification from Occident enable cellular behavior analysis to better engineer T cell therapies for improved cancer treatment.

cell biology↗

bioRxiv2024

Understanding TCR T cell knockout behavior using interpretable machine learning

Genetic perturbation of T cell receptor (TCR) T cells is a promising method to un-lock better TCR T cell performance to create more powerful cancer immunotherapies, but understanding the changes to T cell behavior induced by genetic perturbations remains a challenge. Prior studies have evaluated the effect of different genetic modifications with cytokine production and metabolic activity assays. Live-cell imaging is an inexpensive and robust approach to capture TCR T cell responses to cancer. Most methods to quantify T cell responses in live-cell imaging data use simple approaches to count T cells and cancer cells across time, effectively quantifying how much space in the 2D well each cell type covers, leaving actionable information unexplored. In this study, we characterize changes in TCR T cells interactions with cancer cells from live-cell imaging data using explainable artificial intelligence (AI). We train convolutional neural networks to distinguish behaviors in TCR T cell with CRISPR knock outs of CUL5, RASA2, and a safe harbor control knockout. We use explainable AI to identify specific interaction types that define different knock-out conditions. We find that T cell and cancer cell coverage is a strong marker of TCR T cell modification when comparing similar experimental time points, but differences in cell aggregation characterize CUL5KO and RASA2KO behavior across all time points. Our pipeline for discovery in live-cell imaging data can be used for characterizing complex behaviors in arbitrary live-cell imaging datasets, and we describe best practices for this goal.

cell biology↗

bioRxiv2023

Optimizing the design of spatial genomic studies

Spatially-resolved genomic technologies have shown promise for studying the relationship between the structural arrangement of cells and their functional behavior. While numerous sequencing and imaging platforms exist for performing spatial transcriptomics and spatial proteomics profiling, these experiments remain expensive and labor-intensive. Thus, when performing spatial genomics experiments using multiple tissue slices, there is a need to select the tissue cross sections that will be maximally informative for the purposes of the experiment. In this work, we formalize the problem of experimental design for spatial genomics experiments, which we generalize into a problem class that we call structured batch experimental design. We propose approaches for optimizing these designs in two types of spatial genomics studies: one in which the goal is to construct a spatially-resolved genomic atlas of a tissue and another in which the goal is to localize a region of interest in a tissue, such as a tumor. We demonstrate the utility of these optimal designs, where each slice is a two-dimensional plane, on several spatial genomics datasets.

bioinformatics↗

bioRxiv2022

Expression QTLs in single-cell sequencing data

AO_SCPLOWBSTRACTC_SCPLOWSingle nucleotide polymorphisms (SNPs) are important drivers of gene expression variation and downstream phenotypes including disease risk. Single-cell RNA-sequencing (scRNA-seq) allows an unprecedented exploration of cell-type specific associations between gene expression levels and genotypes, but current methods rely on pseudobulk approaches that use composite expression values across cells and often use summary statistics within cell types, ignoring information across cell types and assuming cell type labels are accurate. Here, we extend our method, telescoping bimodal latent Dirichlet allocation (TBLDA), that identifies covarying genotypes and gene expression values when the matching from samples to cells is not one-to-one in order to allow cell-type label agnostic discovery of eQTLs in noncomposite scRNA-seq data. In particular, we add GPU-compatibility, sparse priors, and amortization to enable fast inference on large-scale scRNA-seq data. We apply single-cell TBLDA (scTBLDA) to 400K cells from 119 individuals with systemic lupus erythematosus and examine properties of features from each modality across the estimated latent factors. We use linked genes and SNPs to identify 205 cis-eQTLS, 66 trans-eQTLs, and 53 cell type proportion QTLs, which we then compare against prior studies of immune-cell eQTLs. Our results demonstrate the ability of scTBLDA to identify genes involved in cell-type specific regulatory processes associated with SNPs in single-cell data.

genomics↗

bioRxiv2022

Linking histology and molecular state across human tissues

Histological imaging and molecular profiling of human tissues both offer information-rich characterizations of biological structure and function. Each of these modalities has been used to characterize the organization and dysregulation of a variety of tissues and cell types. While large-scale studies of each modality in isolation have been conducted, it remains largely unknown the extent to which these two views of a tissue relate to one another. Understanding how cellular states are encoded in cellular morphology would increase the utility and interpretability of imaging data; conversely, understanding the state of the cells within histology images would give deeper insights into the types and states of cells that constitute these tissue samples. To this end, we jointly analyzed 13, 360 human tissue samples with paired bulk gene expression profiles and histology images across 935 donors from the Genotype and Tissue Expression (GTEx) Consortium v8 study. This analysis reveals relationships among gene expression and cellular morphology through shared sources of expression and morphological heterogeneity both within and between tissue types. We describe shared sources of variation including cell-type heterogeneity, sample ischemic time, and donor health and demographics. We find specific correlated effects in both morphology and transcription linked to specific donor characteristics, such as their use of mechanical ventilation. This paired understanding adds value to each data modality on their own by enabling a more precise characterization of the alternative modality in the absence of those data.

genomics↗

bioRxiv2022

A Poisson reduced-rank regression model for association mapping in sequencing data

Single-cell RNA sequencing (scRNA-seq) technologies allow for the study of gene expression in individual cells. Often, it is of interest to understand how transcriptional activity is associated with cell-specific covariates, such as cell type, genotype, or measures of cell health. Traditional approaches for this type of association mapping assume independence between the outcome variables (or genes), and perform a separate regression for each. However, these methods are computationally costly and ignore the substantial correlation structure of gene expression. Furthermore, count-based scRNA-seq data pose challenges for traditional models based on Gaussian assumptions. We aim to resolve these issues by developing a reduced-rank regression model that identifies low-dimensional linear associations between a large number of cell-specific covariates and high-dimensional gene expression readouts. Our probabilistic model uses a Poisson likelihood in order to account for the unique structure of scRNA-seq counts. We demonstrate the performance of our model using simulations, and we apply our model to a scRNA-seq dataset, a spatial gene expression dataset, and a bulk RNA-seq dataset to show its behavior in three distinct analyses. We show that our statistical modeling approach, which is based on reduced-rank regression, captures associations between gene expression and cell- and sample-specific covariates by leveraging low-dimensional representations of transcriptional states.

bioinformatics↗

bioRxiv2022

Alignment of spatial genomics and histology data using deep Gaussian processes

Spatially-resolved genomic technologies have allowed us to study the physical organization of cells and tissues, and promise an understanding of the local interactions between cells. However, it remains difficult to precisely align spatial observations across slices, samples, scales, individuals, and technologies. Here, we propose a probabilistic model that aligns a set of spatially-resolved genomics and histology slices onto a known or unknown common coordinate system into which the samples are aligned both spatially and in terms of the phenotypic readouts (e.g., gene or protein expression levels, cell density, open chromatin regions). Our method consists of a two-layer Gaussian process: the first layer maps the observed samples spatial locations into a common coordinate system, and the second layer maps from the common coordinate system to the observed readouts. Our approach also allows for slices to be mapped to a known template coordinate space if one exists. We show that our registration approach enables complex downstream spatially-aware analyses of spatial genomics data at multiple resolutions that are impossible or inaccurate with unaligned data, including an analysis of variance, differential expression across the z-axis, and association tests across multiple data modalities.

bioinformatics↗

bioRxiv2021

Telescoping bimodal latent Dirichlet allocation to identify expression QTLs across tissues

Expression quantitative trait loci (eQTLs), or single nucleotide polymorphisms (SNPs) that affect average gene expression levels, provide important insights into context-specific gene regulation. Classic eQTL analyses use one-to-one association tests, which test gene-variant pairs individually and ignore correlations induced by gene regulatory networks and linkage disequilibrium. Probabilistic topic models, such as latent Dirichlet allocation, estimate latent topics for a collection of count observations. Prior multi-modal frameworks that bridge genotype and expression data assume matched sample numbers between modalities. However, many data sets have a nested structure where one individual has several associated gene expression samples and a single germline genotype vector. Here, we build a telescoping bimodal latent Dirichlet allocation (TBLDA) framework to learn shared topics across gene expression and genotype data that allows multiple RNA-sequencing samples to correspond to a single individuals genotype. By using raw count data, our model avoids possible adulteration via normalization procedures. Ancestral structure is captured in a genotype-specific latent space, effectively removing it from shared components. Using GTEx v8 expression data across ten tissues and genotype data, we show that the estimated topics capture meaningful and robust biological signal in both modalities, and identify associations within and across tissue types. We identify 53,358 cis-eQTLs and 1,173 trans-eQTLs by conducting eQTL mapping between the most informative features in each topic. Our TBLDA model is able to identify associations using raw sequencing count data when the samples in two separate data modalities are matched one-to-many, as is often the case in biological data.

genomics↗

bioRxiv2021

Differentially methylated regions and methylation QTLs for teen depression and early puberty in the Fragile Families Child Wellbeing Study

AO_SCPLOWBSTRACTC_SCPLOWThe Fragile Families Child Wellbeing Study (FFCWS) is a longitudinal cohort of ethnically diverse and primarily low socioeconomic status children and their families in the U.S. Here, we analyze DNA methylation data collected from 748 FFCWS participants in two waves of this study, corresponding to participant ages 9 and 15. Our primary goal is to leverage the DNA methylation data from these two time points to study methylation associated with two key traits in adolescent health that are over-represented in these data: Early puberty and teen depression. We first identify differentially methylated regions (DMRs) for depression and early puberty. We then identify DMRs for the interaction effects between these two conditions and age by including interaction terms in our regression models to understand how age-related changes in methylation are influenced by depression or early puberty. Next, we identify methylation quantitative trait loci (meQTLs) using genotype data from the participants. We also identify meQTLs with epistatic effects with depression and early puberty. We find enrichment of our interaction meQTLs with functional categories of the genome that contribute to the heritability of co-morbid complex diseases. We replicate our meQTLs in data from the GoDMC study. This work leverages the important focus of the FFCWS data on disadvantaged children to shed light on the methylation states associated with teen depression and early puberty, and on how genetic regulation of methylation is affected in adolescents with these two conditions.

genomics↗