Biology subjects

Ahlmann-Eltze, C.

Publications and source records attributed to Ahlmann-Eltze, C..

6 recordsLinked to original sources

bioRxiv2024

Deep learning-based predictions of gene perturbation effects do not yet outperform simple linear methods

Advanced deep-learning methods, such as foundation models, promise to learn representations of biology that can be employed to predict in silico the outcome of unseen experiments, such as the effect of genetic perturbations on the transcriptomes of human cells. To see whether current models already reach this goal, we benchmarked five foundation models and two other deep learning models against deliberately simplistic linear baselines. For combinatorial perturbations of two genes for which only the individual single perturbations had been seen, we find that the deep learning-based approaches did not perform better than a simple additive model. For perturbations of genes that had not yet been seen, the deep learning-based approaches did not outper-form the baseline of predicting the mean across the training perturbations. We hypothesize that the poor performance is partially because the pre-training data is observational; we show that a simple linear model reliably outperforms all other models when pre-trained on another perturbation dataset. While the promise of deep neural networks for the representation of biological systems and prediction of experimental outcomes is plausible, our work highlights the need for clear setting of objectives and for critical benchmarking to direct research efforts. Contactconstantin.ahlmann@embl.de

bioinformatics↗

bioRxiv2023

Analysis of multi-condition single-cell data with latent embedding multivariate regression

Identifying gene expression differences in heterogeneous tissues across experimental or observational conditions is a fundamental biological task, enabled by single-cell assays such as multi-condition sc-RNA-seq. Current data analysis approaches divide the constituent cells into clusters meant to represent cell types, and identify differentially expressed genes for each cluster. However, such discrete categorization tends to be an unsatisfactory model of the underlying biology. Use of more gradual representations of cell type or cell state promises higher statistical power, better usability and better interpretability. Here, we introduce Latent Embedding Multivariate Regression (LEMUR), a generative model that enables differential expression analysis using a continuous low-dimensional latent space parameterization of cell type and state diversity. It operates without, or before, commitment to discrete categorization. LEMUR (1) integrates data from the different conditions, (2) predicts how each cells gene expression would change as a function of the conditions and its position in latent space, and (3) for each gene, identifies compact neighborhoods of cells with consistent differential expression. Unlike statically defined clusters, these neighborhoods adapt to the underlying gene expression changes. We assess LEMURs performance on a compendium of single-cell datasets and show applications to the identification of tumor subpopulations with distinct drug responses, the interplay between cell state and developmental time in zebrafish embryos, and the discovery of cell state x environment interactions in a spatial single-cell study of plaques in Alzheimers disease. LEMUR is broadly applicable as a first-line analysis approach to multi-condition sc-RNA-seq data. Software availabilityhttps://bioconductor.org/packages/lemur Contactconstantin.ahlmann@embl.de

bioinformatics↗

bioRxiv2021

The S-palmitoylome and DHHC-PAT interactome of Drosophila melanogaster S2R+ cells indicate a high degree of conservation to mammalian palmitoylomes

Protein S-palmitoylation, the addition of a long-chain fatty acid to target proteins, is among the most frequent reversible protein modification in Metazoa, affecting subcellular protein localization, trafficking and protein-protein interactions. S-palmitoylated proteins are abundant in the neuronal system and are associated with neuronal diseases and cancer. Despite the importance of this post-translational modification, it has not been thoroughly studied in the model organism Drosophila melanogaster. Here we present the palmitoylome of Drosophila S2R+ cells, comprising 198 proteins, an estimated 3.5% of expressed genes in these cells. Comparison of orthologs between mammals and Drosophila suggests that S-palmitoylated proteins are more conserved between these distant phyla than non-S-palmitoylated proteins. To identify putative client proteins and interaction partners of the DHHC family of protein acyl-transferases (PATs) we established DHHC-BioID, a proximity biotinylation-based method. In S2R+ cells ectopic expression of the DHHC-PAT dHip14-BioID in combination with Snap24 or an interaction-deficient Snap24- mutant, as a negative control, resulted in biotinylation of Snap24 but not the Snap24-mutant. DHHC-BioID in S2R+ cells using 10 different DHHC-PATs as bait identified 520 putative DHHC-PAT interaction partners of which 48 were S-palmitoylated and are therefore putative DHHC-PAT client proteins. Comparison of putative client protein/DHHC-PAT combinations indicates that CG8314, CG5196, CG5880 and Patsas have a preference for transmembrane proteins, while S-palmitoylated proteins with the Hip14-interaction motif are most enriched by DHHC-BioID variants of approximated and dHip14. Finally, we show that BioID is active in larval and adult Drosophila and that dHip14-BioID rescues dHip14 mutant flies, indicating that DHHC-BioID is non-toxic. In summary we provide the first systematic analysis of a Drosophila palmitoylome. We show that DHHC-BioID is sensitive and specific enough to identify DHHC-PAT client proteins and provide DHHC-PAT assignment for ca. 25% of the S2R+ cell palmitoylome, providing a valuable resource. In addition, we establish DHHC-BioID as a useful concept for the identification of tissue-specific DHHC-PAT interactomes in Drosophila.

biochemistry↗

bioRxiv2021

Transformation and Preprocessing of Single-Cell RNA-Seq Data

The count table, a numeric matrix of genes x cells, is the basic input data structure in the analysis of single-cell RNA-seq data. A common preprocessing step is to adjust the counts for variable sampling efficiency and to transform them so that the variance is similar across the dynamic range. These steps are intended to make subsequent application of generic statistical methods more palatable. Here, we describe four transformation approaches based on the delta method, model residuals, inferred latent expression state, and factor analysis. We compare their strengths and weaknesses and find that the latter three have appealing theoretical properties. However, in benchmarks using simulated and real-world data, it turns out that a rather simple approach, namely, the logarithm with a pseudo-count followed by principal component analysis, performs as well or better than the more sophisticated alternatives. SoftwareThe R package transformGamPoi implementing the delta method- and residuals-based variance-stabilizing transformations is available via Bioconductor. We provide an interactive website to explore the benchmark results at shiny-portal.embl.de/shinyapps/app/08_single-cell_transformation_benchmark. Contactconstantin.ahlmann@embl.de

bioinformatics↗

bioRxiv2020

glmGamPoi: Fitting Gamma-Poisson Generalized Linear Models on Single Cell Count Data

MotivationThe Gamma-Poisson distribution is a theoretically and empirically motivated model for the sampling variability of single cell RNA-sequencing counts (Grun et al., 2014; Townes et al., 2019; Svensson, 2020; Silverman et al., 2018; Hafemeister and Satija, 2019) and an essential building block for analysis approaches including differential expression analysis (Robinson et al., 2010; McCarthy et al., 2012; Anders and Huber, 2010; Love et al., 2014), principal component analysis (Townes et al., 2019) and factor analysis (Risso et al., 2018). Existing implementations for inferring its parameters from data often struggle with the size of single cell datasets, which typically comprise thousands or millions of cells; at the same time, they do not take full advantage of the fact that zero and other small numbers are frequent in the data. These limitations have hampered uptake of the model, leaving room for statistically inferior approaches such as logarithm(-like) transformation. ResultsWe present a new R package for fitting the Gamma-Poisson distribution to data with the characteristics of modern single cell datasets more quickly and more accurately than existing methods. The software can work with data on disk without having to load them into RAM simultaneously. AvailabilityThe package glmGamPoi is available from Bioconductor (since release 3.11) for Windows, macOS, and Linux, and source code is available on GitHub under a GPL-3 license. The scripts to reproduce the results of this paper are available on GitHub as well. Contactconstantin.ahlmann@embl.de

bioinformatics↗

bioRxiv2019

proDA: Probabilistic Dropout Analysis for Identifying Differentially Abundant Proteins in Label-Free Mass Spectrometry

Protein mass spectrometry with label-free quantification (LFQ) is widely used for quantitative proteomics studies. Nevertheless, well-principled statistical inference procedures are still lacking, and most practitioners adopt methods from transcriptomics. These, however, cannot properly treat the principal complication of label-free proteomics, namely many non-randomly missing values. We present proDA, a method to perform statistical tests for differential abundance of proteins. It models missing values in an intensity-dependent probabilistic manner. proDA is based on linear models and thus suitable for complex experimental designs, and boosts statistical power for small sample sizes by using variance moderation. We show that the currently widely used methods based on ad hoc imputation schemes can report excessive false positives, and that proDA not only overcomes this serious issue but also offers high sensitivity. Thus, proDA fills a crucial gap in the toolbox of quantitative proteomics.

bioinformatics↗