Biology subjects

Fischer, D. S.

Publications and source records attributed to Fischer, D. S..

At least 19 recordsLinked to original sources

bioRxiv2025

Notebook-based alignment of human and agentic reasoning in single-cell biology

Agentic AI is increasingly deployed on complex problems, often using chain-of-thought prompting to ground predictions in stepwise reasoning. In biomedical research, assistive agents could make this reasoning accessible to human scientists: for example, intermediate conclusions could be critically evaluated based on shared reasoning and sycophancy - the tendency to affirm a users claims regardless of their validity - could be mitigated. However, this interaction requires an alignment of reasoning between agents and humans that is difficult when multiple data modalities are considered, for example in single-cell biology, where human scientists reason through computational notebooks that contain code, results, and text. To address this issue, we developed kai, an agentic AI that iteratively generates analyses in computational notebooks. We find that kai can flexibly use multiple tools in sequence to solve complex cell type annotation problems. Compared with one-shot generation, kai demonstrates improved robustness against code errors and sycophancy, improved reasoning, and the ability to formulate and address questions on data. kais design is model-agnostic and, therefore, scales directly with advancements in large language models.

bioinformatics↗

bioRxiv2024

Aging disrupts spatiotemporal coordination in the cycling ovary

Throughout the female reproductive lifespan, the ovary completes hundreds of cycles of follicle development, ovulation, and tissue regeneration1-3. These processes rely on the precisely coordinated intricate multicellular interactions across time and space4. How aging disrupts these interactions, leading to an overall decline in reproductive and endocrine functions, remains understudied. To understand the multicellular dynamics that underlie ovarian function and their changes with age, here we use Slide-seq, a near-cellular spatial transcriptomics method, to profile 22 mouse ovaries across the reproductive cycle and chronological age, representing 610,620 near-cellular spots across 69 spatial transcriptomic profiles5,6. We develop a segmentation analysis to identify spatial niches that capture different states of folliculogenesis from static snapshots in situ, allowing us to examine the multicellular dynamics of 358 oocytes, 668 follicles, and 236 corpora lutea. We find that aging disrupts both the spatial organization and temporal coordination of folliculogenesis before the cessation of cycling, which may contribute to the dysregulation of hormone production and signaling. These disruptions are marked by altered immune cell dynamics, inflammatory signaling, and global tissue disorganization that impair the cyclic remodeling required for ovarian function. Our findings reveal how multicellular niches orchestrate ovarian function and demonstrate how age-related breakdown of tissue organization across time and space precedes reproductive decline.

genomics↗

bioRxiv2024

Delineating the Effective Use of Self-Supervised Learning in Single-Cell Genomics

Self-supervised learning (SSL) has emerged as a powerful method for extracting meaningful representations from vast, unlabeled datasets, already transforming computer vision and natural language processing. Similarly, in single-cell genomics (SCG), representation learning is well-recognized for offering insights into complex biological data, even more so by the advent of early foundation model approaches. However, despite these advancements, identifying scenarios in SCG where SSL outperforms traditional supervised or unsupervised learning methods remains a nuanced challenge. Furthermore, selecting the most effective pretext tasks within the SSL framework for SCG is a critical yet unresolved question. Here, we address this gap by adapting and benchmarking SSL techniques in SCG, including masked autoencoders with multiple masking strategies and contrastive learning approaches. Trained on over 20 million cells, this study rigorously examines multiple downstream tasks, including cell type prediction, gene expression reconstruction, cross-modality prediction, and data integration. Our empirical analyses underscore the nuanced role of SSL, namely in transfer learning scenarios leveraging auxiliary data or analyzing novel datasets. Masked autoencoders excel over contrastive methods in SCG, diverging from computer vision trends. Moreover, our findings reveal notable capabilities of SSL in zero-shot cell type prediction and offer insights into its potential benefits in cross-modality prediction and data integration. In summary, we study the application of SSL in SCG, minimizing model bias through simple, fully connected networks, and benchmark SSLs utility across key representation learning scenarios.

bioinformatics↗

bioRxiv2023

Programs, Origins, and Niches of Immunomodulatory Myeloid Cells in Gliomas

Gliomas are incurable malignancies notable for an immunosuppressive microenvironment with abundant myeloid cells whose immunomodulatory properties remain poorly defined. Here, utilizing scRNA-seq data for 183,062 myeloid cells from 85 human tumors, we discover that nearly all glioma-associated myeloid cells express at least one of four immunomodulatory activity programs: Scavenger Immunosuppressive, C1Q Immunosuppressive, CXCR4 Inflammatory, and IL1B Inflammatory. All four programs are present in IDH1 mutant and wild-type gliomas and are expressed in macrophages, monocytes, and microglia whether of blood or resident myeloid cell origins. Integrating our scRNA-seq data with mitochondrial DNA-based lineage tracing, spatial transcriptomics, and organoid explant systems that model peripheral monocyte infiltration, we show that these programs are driven by microenvironmental cues and therapies rather than myeloid cell type, origin, or mutation status. The C1Q Immunosuppressive program is driven by routinely administered dexamethasone. The Scavenger Immunosuppressive program includes ligands with established roles in T-cell suppression, is induced in hypoxic regions, and is associated with immunotherapy resistance. Both immunosuppressive programs are less prevalent in lower-grade gliomas, which are instead enriched for the CXCR4 Inflammatory program. Our study provides a framework to understand immunomodulatory myeloid cells in glioma, and a foundation to develop more effective immunotherapies.

cancer biology↗

bioRxiv2023

Scaling cross-tissue single-cell annotation models

Identifying cellular identities (both novel and well-studied) is one of the key use cases in single-cell transcriptomics. While supervised machine learning has been leveraged to automate cell annotation predictions for some time, there has been relatively little progress both in scaling neural networks to large data sets and in constructing models that generalize well across diverse tissues and biological contexts up to whole organisms. Here, we propose scTab, an automated, feature-attention-based cell type prediction model specific to tabular data, and train it using a novel data augmentation scheme across a large corpus of single-cell RNA-seq observations (22.2 million human cells in total). In addition, scTab leverages deep ensembles for uncertainty quantification. Moreover, we account for ontological relationships between labels in the model evaluation to accommodate for differences in annotation granularity across datasets. On this large-scale corpus, we show that cross-tissue annotation requires nonlinear models and that the performance of scTab scales in terms of training dataset size as well as model size - demonstrating the advantage of scTab over current state-of-the-art linear models in this context. Additionally, we show that the proposed data augmentation schema improves model generalization. In summary, we introduce a de novo cell type prediction model for single-cell RNA-seq data that can be trained across a large-scale collection of curated datasets from a diverse selection of human tissues and demonstrate the benefits of using deep learning methods in this paradigm. Our codebase, training data, and model checkpoints are publicly available at https://github.com/theislab/scTab to further enable rigorous benchmarks of foundation models for single-cell RNA-seq data.

bioinformatics↗

bioRxiv2022

Graph neural networks learn emergent tissue properties from spatial molecular profiles

Tissue phenotypes such as metabolic states, inflammation, and tumor properties are functions of molecular states of cells that constitute the tissue. Recent spatial molecular profiling assays measure tissue architecture motifs in a molecular and often unbiased way and thus can explain some aspects of emergence of these phenotypes. Here, we characterize the ability of graph neural networks to model tissue-level emergent phenotypes based on spatial data by evaluating phenotype prediction across model complexities. First, we show that immune cell dispersion in colorectal tumors, which is known to be predictive of disease outcome, can be captured by graph neural networks. Second, we show that breast cancer tumor classes can be predicted from gene expression alone without spatial information and are thus too simplistic a phenotype to require a complex model of emergence. Third, we show that representation learning approaches for spatial graphs of molecular profiles are limited by overfitting in the prevalent regime of up to 100s of images per study. We address overfitting with within-graph self-supervision and illustrate its promise for tissue representation learning as a constraint for node representations.

bioinformatics↗

bioRxiv2022

Modeling fragment counts improves single-cell ATAC-seq analysis

Single-cell ATAC-sequencing (scATAC-seq) coverage in regulatory regions is typically binarized as an indicator of open chromatin. However, the implications of scATAC-seq data binarization have not systematically been assessed. Here, we show that the goodness-of-fit of existing models and their applications, including clustering, cell type identification, and batch integration, are improved by a quantitative treatment of the fragment counts. These results have immediate implications for scATAC-seq analysis.

bioinformatics↗

bioRxiv2021

Learning cell communication from spatial graphs of cells

Tissue niches are sources of cellular variation and key to understanding both single-cell and tissue phenotypes. The interaction of a cell with its niche can be described through cell communication events. These events cannot be directly observed in molecular profiling assays of single cells and have to be inferred. However, computational models of cell communication and variance attribution defined on data from dissociated tissues suffer from multiple limitations with respect to their ability to define and to identify communication events. We address these limitations using spatial molecular profiling data with node-centric expression modeling (NCEM), a computational method based on graph neural networks which reconciles variance attribution and communication modeling in a single model of tissue niches. We use these models in varying complexity across spatial assays, such as immunohistochemistry and MERFISH, and biological systems to demonstrate that the statistical cell-cell dependencies discovered by NCEM are plausible signatures of known molecular processes underlying cell communication. We identify principles of tissue organisation as cell communication events across multiple datasets using interpretation mechanisms. In the primary motor cortex, we found gene expression variation that is due to niche composition variation across cortical depth. Using the same approach, we also identified niche-dependent cell state variation in CD8 T cells from inflamed colon and colorectal cancer. Finally, we show that NCEMs can be extended to mixed models of explicit cell communication events and latent intrinsic sources of variation in conditional variational autoencoders to yield holistic models of cellular variation in spatial molecular profiling data. Altogether, this graphical model of cellular niches is a step towards understanding emergent tissue phenotypes.

bioinformatics↗

bioRxiv2021

Squidpy: a scalable framework for spatial single cell analysis

Spatial omics data are advancing the study of tissue organization and cellular communication at an unprecedented scale. Here, we present Squidpy, a Python framework that brings together tools from omics and image analysis to enable scalable description of spatial molecular data, such as transcriptome or multivariate proteins. Squidpy provides both infrastructure and numerous analysis methods that allow to efficiently store, manipulate and interactively visualize spatial omics data.

bioinformatics↗

bioRxiv2020

Ultra-high sensitivity mass spectrometry quantifies single-cell proteome changes upon perturbation

Single-cell technologies are revolutionizing biology but are today mainly limited to imaging and deep sequencing1-3. However, proteins are the main drivers of cellular function and in-depth characterization of individual cells by mass spectrometry (MS)-based proteomics would thus be highly valuable and complementary4,5. Chemical labeling-based single-cell approaches introduce hundreds of cells into the MS, but direct analysis of single cells has not yet reached the necessary sensitivity, robustness and quantitative accuracy to answer biological questions6,7. Here, we develop a robust workflow combining miniaturized sample preparation, very low flow-rate chromatography and a novel trapped ion mobility mass spectrometer, resulting in a more than ten-fold improved sensitivity. We accurately and robustly quantify proteomes and their changes in single, FACS-isolated cells. Arresting cells at defined stages of the cell cycle by drug treatment retrieves expected key regulators such as CDK2NA, the E2 ubiquitin ligase UBE2S, DNA topoisomerases TOP2A/B and the chromatin regulator HMGA1. Furthermore, it highlights potential novel ones and allows cell phase prediction. Comparing the variability in more than 430 single-cell proteomes to transcriptome data revealed a stable core proteome despite perturbation, while the transcriptome appears volatile. This emphasizes substantial regulation of translation and sets the stage for its elucidation at the single cell level. Our technology can readily be applied to ultra-high sensitivity analyses of tissue material8, posttranslational modifications and small molecule studies to gain unprecedented insights into cellular heterogeneity in health and disease.

systems biology↗

bioRxiv2020

Sfaira accelerates data and model reuse in single cell genomics

Exploratory analysis of single-cell RNA-seq data sets is currently based on statistical and machine learning models that are adapted to each new data set from scratch. A typical analysis workflow includes a choice of dimensionality reduction, selection of clustering parameters, and mapping of prior annotation. These steps typically require several iterations and can take up significant time in many single-cell RNA-seq projects. Here, we introduce sfaira, which is a single-cell data and model zoo which houses data sets as well as pre-trained models. The data zoo is designed to facilitate the fast and easy contribution of data sets, interfacing to a large community of data providers. Sfaira currently includes 233 data sets across 45 organs and 3.1 million cells in both human and mouse. Using these data sets we have trained eight different example model classes, such as autoencoders and logistic cell type predictors: The infrastructure of sfaira is model agnostic and allows training und usage of many previously published models. Sfaira directly aids in exploratory data analysis by replacing embedding and cell type annotation workflows with end-to-end pre-trained parametric models. As further example use cases for sfaira, we demonstrate the extraction of gene-centric data statistics across many tissues, improved usage of cell type labels at different levels of coarseness, and an application for learning interpretable models through data regularization on extremely diverse data sets.

bioinformatics↗

bioRxiv2020

Identification and characterization of distinct murine brown adipocyte lineages

Brown adipose tissue (BAT) plays an important role in the regulation of body weight and glucose homeostasis. While increasing evidence supports white adipose tissue heterogeneity, little is known about heterogeneity within murine BAT. Using single cell RNA sequencing of the stromal vascular fraction of murine BAT and analysis of 67 brown preadipocyte and adipocyte clones we unravel heterogeneity within brown preadipocytes. Statistical analysis of gene expression profiles from these clones identifies markers distinguishing brown adipocyte lineages. We confirm the presence of distinct brown adipocyte populations in vivo using three identified markers; Eif5, Tcf25, and Bin1. Functionally, we demonstrate that loss of Bin1 enhances UCP1 expression and mitochondrial respiration, suggesting that Bin1 marks a dormant brown adipocyte type. The existence of multiple brown adipocyte lineages suggests distinct functional properties of BAT depending on its cellular composition, with potentially distinct function in thermogenesis and the regulation of whole body energy homeostasis.

cell biology↗

bioRxiv2020

Integrated analyses of single-cell atlases reveal age, gender, and smoking status associations with cell type-specific expression of mediators of SARS-CoV-2 viral entry and highlights inflammatory programs in putative target cells

The COVID-19 pandemic, caused by the novel coronavirus SARS-CoV-2, creates an urgent need for identifying molecular mechanisms that mediate viral entry, propagation, and tissue pathology. Cell membrane bound angiotensin-converting enzyme 2 (ACE2) and associated proteases, transmembrane protease serine 2 (TMPRSS2) and Cathepsin L (CTSL), were previously identified as mediators of SARS-CoV2 cellular entry. Here, we assess the cell type-specific RNA expression of ACE2, TMPRSS2, and CTSL through an integrated analysis of 107 single-cell and single-nucleus RNA-Seq studies, including 22 lung and airways datasets (16 unpublished), and 85 datasets from other diverse organs. Joint expression of ACE2 and the accessory proteases identifies specific subsets of respiratory epithelial cells as putative targets of viral infection in the nasal passages, airways, and alveoli. Cells that co-express ACE2 and proteases are also identified in cells from other organs, some of which have been associated with COVID-19 transmission or pathology, including gut enterocytes, corneal epithelial cells, cardiomyocytes, heart pericytes, olfactory sustentacular cells, and renal epithelial cells. Performing the first meta-analyses of scRNA-seq studies, we analyzed 1,176,683 cells from 282 nasal, airway, and lung parenchyma samples from 164 donors spanning fetal, childhood, adult, and elderly age groups, associate increased levels of ACE2, TMPRSS2, and CTSL in specific cell types with increasing age, male gender, and smoking, all of which are epidemiologically linked to COVID-19 susceptibility and outcomes. Notably, there was a particularly low expression of ACE2 in the few young pediatric samples in the analysis. Further analysis reveals a gene expression program shared by ACE2+TMPRSS2+ cells in nasal, lung and gut tissues, including genes that may mediate viral entry, subtend key immune functions, and mediate epithelial-macrophage cross-talk. Amongst these are IL6, its receptor and co-receptor, IL1R, TNF response pathways, and complement genes. Cell type specificity in the lung and airways and smoking effects were conserved in mice. Our analyses suggest that differences in the cell type-specific expression of mediators of SARS-CoV-2 viral entry may be responsible for aspects of COVID-19 epidemiology and clinical course, and point to putative molecular pathways involved in disease susceptibility and pathogenesis.

bioinformatics↗

bioRxiv2020

Automatic identification of relevant genes from low-dimensional embeddings of single cell RNAseq data

Dimensionality reduction is a key step in the analysis of single-cell RNA sequencing data and produces a low-dimensional embedding for visualization and as a calculation base for downstream analysis. Nonlinear techniques are most suitable to handle the intrinsic complexity of large, heterogeneous single cell data. With no linear relation between genes and embedding however, there is no way to extract the identity of genes most relevant for any cells position in the low-dimensional embedding, and thus the underlying process. In this paper, we introduce the concepts of global and local gene relevance to compute an equivalent of principal component analysis loadings for non-linear low-dimensional embeddings. While global gene relevance identifies drivers of the overall embedding, local gene relevance singles out genes that change in small, possibly rare subsets of cells. We apply our method to single-cell RNAseq datasets from different experimental protocols and to different low dimensional embedding techniques, shows our methods versatility to identify key genes for a variety of biological processes. To ensure reproducibility and ease of use, our method is released as part of destiny 3.0, a popular R package for building diffusion maps from single-cell transcriptomic data. It is readily available through Bioconductor.

bioinformatics↗

bioRxiv2020

Cell-type specific impact of glucocorticoid receptor activation on the developing brain

A fine-tuned balance of glucocorticoid receptor (GR) activation is essential for organ formation, with disturbances influencing health outcomes. Excess GR-activation in utero has been linked to brain-related negative outcomes, with unclear underlying mechanisms, especially regarding cell-type specific effects. To address this, we used an in vitro model of fetal human brain, induced pluripotent-stem-cell-derived cerebral organoids, and mapped GR-activation effects using single-cell transcriptomics across development. Interestingly, neurons showed targeted regulation of differentiation- and maturation-related transcripts, suggesting a delay of these processes upon GR-activation. Uniquely in neurons, differentially-expressed transcripts were significantly enriched for genes associated with behavior-related phenotypes and disorders. This suggests that aberrant GR-activation could impact proper neuronal maturation, leading to increased disease susceptibility, through neurodevelopmental processes at the interface of genetic susceptibility and environmental exposure.

neuroscience↗

bioRxiv2019

Predicting antigen-specificity of single T-cells based on TCR CDR3 regions

It has recently become possible to assay T-cell specificity with respect to large sets of antigens as well as T-cell receptor sequence in high-throughput single-cell experiments. We propose multiple sequence-data specific deep learning approaches to impute TCR to epitope specificity to reduce the complexity of new experiments. We found that models that treat antigens as categorical variables outperform those which model the TCR and epitope sequence jointly. Moreover, we show that variability in single-cell immune repertoire screens can be mitigated by modeling cell-specific covariates.

bioinformatics↗

bioRxiv2019

epiScanpy: integrated single-cell epigenomic analysis

Epigenetic single-cell measurements reveal a layer of regulatory information not accessible to single-cell transcriptomics, however single-cell-omics analysis tools mainly focus on gene expression data. To address this issue, we present epiScanpy, a computational framework for the analysis of single-cell DNA methylation and single-cell ATAC-seq data. EpiScanpy makes the many existing RNA-seq workflows from scanpy available to large-scale single-cell data from other -omics modalities. We introduce and compare multiple feature space constructions for epigenetic data and show the feasibility of common clustering, dimension reduction and trajectory learning techniques. We benchmark epiScanpy by interrogating different single-cell brain mouse atlases of DNA methylation, ATAC-seq and transcriptomics. We find that differentially methylated and differentially open markers between cell clusters enrich transcriptome-based cell type labels by orthogonal epigenetic information.

bioinformatics↗

bioRxiv2019

MPRAnalyze - A statistical framework for Massively Parallel Reporter Assays

Massively parallel reporter assays (MPRAs) are a technique that enables testing thousands of regulatory DNA sequences and their variants in a single, quantitative experiment. Despite growing popularity, there is lack of statistical methods that account for the different sources of uncertainty inherent to these assays, thus effectively leveraging their promise. Development of such methods could help enhance our ability to identify regulatory sequences in the genome, understand their function under various setting, and ultimately gain a better understanding of how the regulatory code and its alteration lead to phenotypic consequence. Here we present MPRAnalyze: a statistical framework dedicated to analyzing MPRA count data. MPRAnalyze addresses the major questions that are posed in the context of MPRA experiments: estimating the magnitude of the effect of a regulatory sequence in a single condition setting, and comparing differential activity of regulatory sequences across multiple conditions. The framework uses a nested construction of generalized linear models to account for uncertainty in both DNA and RNA observations, controls for various sources of unwanted variation, and incorporates negative controls for robust hypothesis testing, thereby providing clear quantitative answers in complex experimental settings. We demonstrate the robustness, accuracy and applicability of MPR-Analyze on simulated data and published data sets and compare it against the existing analysis methodologies. MPRAnalyze is implemented as an R package and is publicly available through Bioconductor [1].

bioinformatics↗