Britta Velten
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Britta Velten has a total of 164 co-authors. Here are the top co-authors by number of shared publications:
- Oliver Stegle
Genome Biology Unit, European Molecular Biology Laboratory (EMBL), Heidelberg, Germany
10 shared publications - Wolfgang Huber
Genome Biology Unit, European Molecular Biology Laboratory (EMBL), Heidelberg, Germany
9 shared publications - Jana Muriel Braunger
Heidelberg University, Centre for Organismal Studies
7 shared publications - Junyan Lu
Medical Faculty Heidelberg, Heidelberg University, Heidelberg, Germany
6 shared publications - Damien Arnol
European Molecular Biology Laboratory, European Bioinformatics Institute (EMBL-EBI), Cambridge, UK
4 shared publications - Ricard Argelaguet
Epigenetics Programme, Babraham Institute, Cambridge CB22 3AT, UK
4 shared publications
Latest Publications
- Open Access
- 14 November 2026
- Published by Springer Nature in BMC Bioinformatics
Comparison of interventional causal structure learning algorithms for gene regulatory network inference
Background: Causal structure learning offers a promising approach to studying gene regulation in cells, aiming to provide deeper mechanistic insights than purely association-based methods. Theoretical groundwork indicating that interventions improve identifiability of causal structure motivates the use of causal structure learning methods in scenarios with interventional information. Results: This benchmark investigates the ability of existing causal structure learning algorithms to leverage the information revealed by targeted interventions to infer gene regulatory networks (GRNs). In this study synthetic single-cell perturbation data and experimental data from a clustered regularly interspaced short palindromic repeats (CRISPR) screen are leveraged, and a suite of causal structure learning algorithms is evaluated on metrics tailored to synthetic ground truth and real biological data respectively. On synthetic data, accurate recovery of GRNs was achieved under favourable conditions: strong interventions, large sample sizes, and low measurement noise. For experimental data, no reliable ground truth is available, so inferred GRNs were assessed indirectly, using observed perturbation effects and curated regulatory annotations. Here, performance remained unreliable, with technical and biological noise and algorithmic scalability as limiting factors. This highlights a current gap between theoretical potential and practical application of causal structure learning for GRNs. Conclusions: The benchmark provides insight into algorithm strengths and limitations and offers groundwork for further methodological development. We also provide an accessible software package for applying modern causal structure learning to custom datasets and thereby foster future exploration of its potential.
- 14 November 2026
- Published by Springer Nature in BMC Bioinformatics
Open AccessComparison of interventional causal structure learning algorithms for gene regulatory network inference
Background: Causal structure learning offers a promising approach to studying gene regulation in cells, aiming to provide deeper mechanistic insights than purely association-based methods. Theoretical groundwork indicating that interventions improve identifiability of causal structure motivates the use of causal structure learning methods in scenarios with interventional information. Results: This benchmark investigates the ability of existing causal structure learning algorithms to leverage the information revealed by targeted interventions to infer gene regulatory networks (GRNs). In this study synthetic single-cell perturbation data and experimental data from a clustered regularly interspaced short palindromic repeats (CRISPR) screen are leveraged, and a suite of causal structure learning algorithms is evaluated on metrics tailored to synthetic ground truth and real biological data respectively. On synthetic data, accurate recovery of GRNs was achieved under favourable conditions: strong interventions, large sample sizes, and low measurement noise. For experimental data, no reliable ground truth is available, so inferred GRNs were assessed indirectly, using observed perturbation effects and curated regulatory annotations. Here, performance remained unreliable, with technical and biological noise and algorithmic scalability as limiting factors. This highlights a current gap between theoretical potential and practical application of causal structure learning for GRNs. Conclusions: The benchmark provides insight into algorithm strengths and limitations and offers groundwork for further methodological development. We also provide an accessible software package for applying modern causal structure learning to custom datasets and thereby foster future exploration of its potential.
- Open Access
- 7 July 2026
- Published by Springer Nature in Communications Chemistry
- Vol. 9 (1)
msBayesImpute as a versatile framework for addressing missing values in biomedical mass spectrometry proteomics data
Advancements in mass spectrometry (MS) technologies have significantly improved the ability to quantify proteins and analyse their modifications. However, MS-based proteomics datasets frequently encounter missing values due to a complex interplay of missing at random (MAR) and missing not at random (MNAR) mechanisms. Such missing data can result in information loss and biased outcomes in data pre-processing, as well as subsequent analyses and interpretations. Few approaches effectively address both MAR and MNAR, and those that do often necessitate manual tuning of mixture percentages between them or rely on two-group experimental designs. Therefore, we developed msBayesImpute, an innovative computational method that integrates Bayesian factorization with probabilistic dropout models. We evaluated msBayesImpute against several popular imputation methods using both simulated missing values and those generated through a dilution series experiment on samples from lung cancer patients. Our comprehensive benchmark demonstrated superior performance in reconstructing missing values, estimating normalization factors, identifying differentially expressed proteins and predicting outcomes with machine learning models across varying levels of missingness and sample sizes. Notably, msBayesImpute does not require predefined experimental designs and is scalable to large-scale studies. This versatility positions msBayesImpute as an effective and robust tool for enhancing the utility of MS datasets in biological research.
- 7 July 2026
- Published by Springer Nature in Communications Chemistry
- Vol. 9 (1)
Open AccessmsBayesImpute as a versatile framework for addressing missing values in biomedical mass spectrometry proteomics data
Advancements in mass spectrometry (MS) technologies have significantly improved the ability to quantify proteins and analyse their modifications. However, MS-based proteomics datasets frequently encounter missing values due to a complex interplay of missing at random (MAR) and missing not at random (MNAR) mechanisms. Such missing data can result in information loss and biased outcomes in data pre-processing, as well as subsequent analyses and interpretations. Few approaches effectively address both MAR and MNAR, and those that do often necessitate manual tuning of mixture percentages between them or rely on two-group experimental designs. Therefore, we developed msBayesImpute, an innovative computational method that integrates Bayesian factorization with probabilistic dropout models. We evaluated msBayesImpute against several popular imputation methods using both simulated missing values and those generated through a dilution series experiment on samples from lung cancer patients. Our comprehensive benchmark demonstrated superior performance in reconstructing missing values, estimating normalization factors, identifying differentially expressed proteins and predicting outcomes with machine learning models across varying levels of missingness and sample sizes. Notably, msBayesImpute does not require predefined experimental designs and is scalable to large-scale studies. This versatility positions msBayesImpute as an effective and robust tool for enhancing the utility of MS datasets in biological research.
- Open Access
Statistical tests for bivariate spatial association across multi-omics data with disjoint coordinates
Spatial biology has entered a new era of multimodal profiling, with multiple, high-dimensional spatial omics types being measured on consecutive tissue slices, or co-assayed on the same slice. Interest then lies in statistical testing for spatial association between the features of the different modalities, to gain insight in biological processes. One major challenge is the multitude of bivariate combinations, leading to high computational demands. Another difficulty is the difference in spatial resolution between technologies, implying no one-to-one matching between the measurement spots of the two modalities, even after alignment. As a result, common statistical measures such as joint distributions and correlations are not defined, and tests need to rely on spatial vicinity only. Moreover, we argue that many existing bivariate association tests address an inappropriate null hypothesis, or make inappropriate assumptions, both implying absence of spatial autocorrelation in any of the features and leading to misleading conclusions. As a remedy, we modify tests for the detection of spatially variable genes (Moran's I, Gaussian processes and generalized additive models (splines)) to derive bivariate tests across modalities with non-overlapping coordinate sets and provide variance estimators that do account for spatial autocorrelation. We develop inference methods for single sections as well as for replicated experiments with multiple sections, and compare their performance in nonparametric and parametric simulations. Finally, we apply the newly developed methods to two co-assayed spatial transcriptomics and metabolomics datasets from mouse and human. The full suite of tests is available from github.com/sthawinke/sbivar as the R-package sbivar.
Open AccessStatistical tests for bivariate spatial association across multi-omics data with disjoint coordinates
Spatial biology has entered a new era of multimodal profiling, with multiple, high-dimensional spatial omics types being measured on consecutive tissue slices, or co-assayed on the same slice. Interest then lies in statistical testing for spatial association between the features of the different modalities, to gain insight in biological processes. One major challenge is the multitude of bivariate combinations, leading to high computational demands. Another difficulty is the difference in spatial resolution between technologies, implying no one-to-one matching between the measurement spots of the two modalities, even after alignment. As a result, common statistical measures such as joint distributions and correlations are not defined, and tests need to rely on spatial vicinity only. Moreover, we argue that many existing bivariate association tests address an inappropriate null hypothesis, or make inappropriate assumptions, both implying absence of spatial autocorrelation in any of the features and leading to misleading conclusions. As a remedy, we modify tests for the detection of spatially variable genes (Moran's I, Gaussian processes and generalized additive models (splines)) to derive bivariate tests across modalities with non-overlapping coordinate sets and provide variance estimators that do account for spatial autocorrelation. We develop inference methods for single sections as well as for replicated experiments with multiple sections, and compare their performance in nonparametric and parametric simulations. Finally, we apply the newly developed methods to two co-assayed spatial transcriptomics and metabolomics datasets from mouse and human. The full suite of tests is available from github.com/sthawinke/sbivar as the R-package sbivar.
- Open Access
Genome-scale mapping of variant, enhancer and gene function in primary human CD4+ T cells
CD4+ T cells harbor a disproportionate enrichment of immune disease risk loci and represent the primary cellular context for immune disease biology, yet the genes and regulatory programs these variants affect remain largely unknown. We combined targeted Perturb-seq of 1,032 cis -regulatory elements (CREs) overlapping 4,724 variants across 14 immune diseases with genome-wide Perturb-seq of all expressed genes in primary human CD4+ T cells, spanning 4.1 million cells. We identified 626 CRE-gene pairs, and connected CRE targets to downstream regulatory cascades. At the TYK2 and DEXI /CLEC16A loci, we resolved target genes and linked noncoding variants to inflammatory and metabolic programs. Across diseases, we revealed that dispersed variants converged on shared and disease-specific programs. Our work provides a framework for tracing variant-to-CRE-to-gene-to-network in disease-relevant primary cells.
Open AccessGenome-scale mapping of variant, enhancer and gene function in primary human CD4+ T cells
CD4+ T cells harbor a disproportionate enrichment of immune disease risk loci and represent the primary cellular context for immune disease biology, yet the genes and regulatory programs these variants affect remain largely unknown. We combined targeted Perturb-seq of 1,032 cis -regulatory elements (CREs) overlapping 4,724 variants across 14 immune diseases with genome-wide Perturb-seq of all expressed genes in primary human CD4+ T cells, spanning 4.1 million cells. We identified 626 CRE-gene pairs, and connected CRE targets to downstream regulatory cascades. At the TYK2 and DEXI /CLEC16A loci, we resolved target genes and linked noncoding variants to inflammatory and metabolic programs. Across diseases, we revealed that dispersed variants converged on shared and disease-specific programs. Our work provides a framework for tracing variant-to-CRE-to-gene-to-network in disease-relevant primary cells.
- Open Access
Comparison of Interventional Causal Structure Learning Algorithms for Gene Regulatory Network Inference
Background: Causal structure learning offers a promising approach to studying gene regulation in cells, aiming to provide deeper mechanistic insights than purely association-based methods. Theoretical groundwork indicating that interventions improve identifiability of causal structure motivates the use of causal structure learning methods in scenarios with interventional information. Results: This benchmark investigates the ability of existing causal structure learning algorithms to leverage the information revealed by targeted interventions to infer gene regulatory networks (GRNs). In this study both synthetic and experimental single-cell CRISPR perturbation data is leveraged, and a suite of causal structure learning algorithms, is evaluated on metrics tailored to synthetic ground truth and real biological data respectively. On synthetic data, accurate recovery of GRNs is achieved under favourable conditions: strong interventions, large sample sizes, and low measurement noise. However, on real data, performance remains unreliable, limited by technical and biological noise, as well as algorithmic scalability. This highlights a current gap between theoretical potential and practical application of causal structure learning for GRNs. Conclusions: The benchmark provides insight into algorithm strengths and limitations and offers groundwork for further methodological development. We also provide an accessible software package to leverage modern causal structure learning on custom datasets and thereby foster future exploration of the potential of causal structure learning.
Open AccessComparison of Interventional Causal Structure Learning Algorithms for Gene Regulatory Network Inference
Background: Causal structure learning offers a promising approach to studying gene regulation in cells, aiming to provide deeper mechanistic insights than purely association-based methods. Theoretical groundwork indicating that interventions improve identifiability of causal structure motivates the use of causal structure learning methods in scenarios with interventional information. Results: This benchmark investigates the ability of existing causal structure learning algorithms to leverage the information revealed by targeted interventions to infer gene regulatory networks (GRNs). In this study both synthetic and experimental single-cell CRISPR perturbation data is leveraged, and a suite of causal structure learning algorithms, is evaluated on metrics tailored to synthetic ground truth and real biological data respectively. On synthetic data, accurate recovery of GRNs is achieved under favourable conditions: strong interventions, large sample sizes, and low measurement noise. However, on real data, performance remains unreliable, limited by technical and biological noise, as well as algorithmic scalability. This highlights a current gap between theoretical potential and practical application of causal structure learning for GRNs. Conclusions: The benchmark provides insight into algorithm strengths and limitations and offers groundwork for further methodological development. We also provide an accessible software package to leverage modern causal structure learning on custom datasets and thereby foster future exploration of the potential of causal structure learning.
- Open Access
- 1 December 2025
- Published by Elsevier in Cell Genomics
- Vol. 6 (2) , 101076
A genome-scale single-cell CRISPRi map of trans gene regulation across human pluripotent stem cell lines
- 1 December 2025
- Published by Elsevier in Cell Genomics
- Vol. 6 (2) , 101076
Open AccessA genome-scale single-cell CRISPRi map of trans gene regulation across human pluripotent stem cell lines
- Open Access
msBayesImpute: A Versatile Framework for Addressing Missing Values in Biomedical Mass Spectrometry Proteomics Data
Advancements in mass spectrometry (MS) technologies have significantly improved the ability to quantify proteins and analyse their modifications. However, MS-based proteomics datasets frequently encounter missing values due to a complex interplay of missing at random (MAR) and missing not at random (MNAR) mechanisms. If unaddressed, such missing data can result in information loss and biased outcomes in data pre-processing, such as normalisation, as well as subsequent analyses and interpretations. Few approaches effectively address both MAR and MNAR, and those that do often necessitate manual tuning of mixture percentages between them or rely on two-group experimental designs. To enhance the handling of missing values, we developed msBayesImpute, an innovative computational method that integrates Bayesian factorization with probabilistic dropout models. We evaluated msBayesImpute against several popular imputation methods using both simulated missing values and those generated through a dilution series experiment on samples from lung cancer patients. Our comprehensive benchmark demonstrated superior performance in reconstructing missing values, estimating normalization factors, and identifying differentially expressed proteins across varying levels of missingness. Notably, msBayesImpute does not require predefined experimental designs and is scalable to large-scale studies. This versatility positions msBayesImpute as an effective and robust tool for enhancing the utility of MS datasets in biological research.
Open AccessmsBayesImpute: A Versatile Framework for Addressing Missing Values in Biomedical Mass Spectrometry Proteomics Data
Advancements in mass spectrometry (MS) technologies have significantly improved the ability to quantify proteins and analyse their modifications. However, MS-based proteomics datasets frequently encounter missing values due to a complex interplay of missing at random (MAR) and missing not at random (MNAR) mechanisms. If unaddressed, such missing data can result in information loss and biased outcomes in data pre-processing, such as normalisation, as well as subsequent analyses and interpretations. Few approaches effectively address both MAR and MNAR, and those that do often necessitate manual tuning of mixture percentages between them or rely on two-group experimental designs. To enhance the handling of missing values, we developed msBayesImpute, an innovative computational method that integrates Bayesian factorization with probabilistic dropout models. We evaluated msBayesImpute against several popular imputation methods using both simulated missing values and those generated through a dilution series experiment on samples from lung cancer patients. Our comprehensive benchmark demonstrated superior performance in reconstructing missing values, estimating normalization factors, and identifying differentially expressed proteins across varying levels of missingness. Notably, msBayesImpute does not require predefined experimental designs and is scalable to large-scale studies. This versatility positions msBayesImpute as an effective and robust tool for enhancing the utility of MS datasets in biological research.
- Open Access
A genome-scale single cell CRISPRi map of trans gene regulation across human pluripotent stem cell lines
Population-scale resources of genetic, molecular, and cellular information form the basis for understanding human genomes, charting the heritable basis of disease, and tracing the effects of mutations. Pooled perturbation assays applied to cellular models, probing the effect of many perturbations coupled with an scRNA-seq readout (Perturb-seq), are especially potent references for interpreting disease-linked mutations or gene expression changes. However, the utility of existing maps has been limited by the comprehensiveness of perturbations possible, and the relevance of their cell line context. Here, we present the first genome-scale CRISPR interference (CRISPRi) perturbation map with single-cell RNA sequencing readout across many human genetic backgrounds in human pluripotent cells. To do so, we establish large-scale CRISPRi screening in human induced pluripotent stem cells from healthy donors, using over 20,000 guide RNAs to target 7,226 genes across 34 cell lines from 26 genetic backgrounds, and gather expression data from nearly 2 million cells. We comprehensively map trans expression changes induced by the target knockdowns, which complement co-expression patterns in unperturbed cells and facilitate the functional annotation of target genes to biological processes and complexes. Consistency of targeting protein complex members point to protein complexes as a nexus for aggregating transcriptional variation, revealing novel interaction partners. We characterise variation in perturbation effects across donors, with expression quantitative trait loci linked to higher genetic modulation of perturbation effects but overall low replication of trans effects due to knockdown of their corresponding cis regulators. This study pioneers population-scale CRISPR perturbations with single cell readouts that will fuel foundation models for the future of effective modulation of cellular disease phenotypes.
Open AccessA genome-scale single cell CRISPRi map of trans gene regulation across human pluripotent stem cell lines
Population-scale resources of genetic, molecular, and cellular information form the basis for understanding human genomes, charting the heritable basis of disease, and tracing the effects of mutations. Pooled perturbation assays applied to cellular models, probing the effect of many perturbations coupled with an scRNA-seq readout (Perturb-seq), are especially potent references for interpreting disease-linked mutations or gene expression changes. However, the utility of existing maps has been limited by the comprehensiveness of perturbations possible, and the relevance of their cell line context. Here, we present the first genome-scale CRISPR interference (CRISPRi) perturbation map with single-cell RNA sequencing readout across many human genetic backgrounds in human pluripotent cells. To do so, we establish large-scale CRISPRi screening in human induced pluripotent stem cells from healthy donors, using over 20,000 guide RNAs to target 7,226 genes across 34 cell lines from 26 genetic backgrounds, and gather expression data from nearly 2 million cells. We comprehensively map trans expression changes induced by the target knockdowns, which complement co-expression patterns in unperturbed cells and facilitate the functional annotation of target genes to biological processes and complexes. Consistency of targeting protein complex members point to protein complexes as a nexus for aggregating transcriptional variation, revealing novel interaction partners. We characterise variation in perturbation effects across donors, with expression quantitative trait loci linked to higher genetic modulation of perturbation effects but overall low replication of trans effects due to knockdown of their corresponding cis regulators. This study pioneers population-scale CRISPR perturbations with single cell readouts that will fuel foundation models for the future of effective modulation of cellular disease phenotypes.
- Open Access
- 2 September 2024
- Published by Oxford University Press (OUP) in Bioinformatics
- Vol. 40 (9)
Guide assignment in single-cell CRISPR screens using crispat
Motivation: Pooled single-cell CRISPR screens have emerged as a powerful tool in functional genomics to probe the effect of genetic interventions at scale. A crucial step in the analysis of the resulting data is the assignment of cells to gRNAs corresponding to a specific genetic intervention. However, this step is challenging due to a lack of systematic benchmarks and accessible software to apply and compare different guide assignment strategies. To address this, we here propose crispat (CRISPR guide assignment tool), a Python package to facilitate the choice of a suitable guide assignment strategy for single-cell CRISPR screens. Results: We demonstrate the package on four single-cell CRISPR interference screens at low multiplicity of infection from two studies, where crispat identifies strong differences in the number of assigned cells, downregulation of the target genes and number of discoveries across different guide assignment strategies, highlighting the need for a suitable guide assignment strategy to obtain optimal power in single-cell CRISPR screens. Availability and implementation: crispat is implemented in python, the source code, installation instructions and tutorials can be found at https://github.com/velten-group/crispat and it can be installed from PyPI (https://pypi.org/project/crispat/). Code to reproduce all findings in this paper is available at https://github.com/velten-group/crispat_analysis, as well as at https://zenodo.org/records/13373265.
- 2 September 2024
- Published by Oxford University Press (OUP) in Bioinformatics
- Vol. 40 (9)
Open AccessGuide assignment in single-cell CRISPR screens using crispat
Motivation: Pooled single-cell CRISPR screens have emerged as a powerful tool in functional genomics to probe the effect of genetic interventions at scale. A crucial step in the analysis of the resulting data is the assignment of cells to gRNAs corresponding to a specific genetic intervention. However, this step is challenging due to a lack of systematic benchmarks and accessible software to apply and compare different guide assignment strategies. To address this, we here propose crispat (CRISPR guide assignment tool), a Python package to facilitate the choice of a suitable guide assignment strategy for single-cell CRISPR screens. Results: We demonstrate the package on four single-cell CRISPR interference screens at low multiplicity of infection from two studies, where crispat identifies strong differences in the number of assigned cells, downregulation of the target genes and number of discoveries across different guide assignment strategies, highlighting the need for a suitable guide assignment strategy to obtain optimal power in single-cell CRISPR screens. Availability and implementation: crispat is implemented in python, the source code, installation instructions and tutorials can be found at https://github.com/velten-group/crispat and it can be installed from PyPI (https://pypi.org/project/crispat/). Code to reproduce all findings in this paper is available at https://github.com/velten-group/crispat_analysis, as well as at https://zenodo.org/records/13373265.
- Open Access
Guide assignment in single-cell CRISPR screens using crispat
Pooled single cell CRISPR screens have emerged as a powerful tool in functional genomics to probe the effect of genetic interventions at scale. A crucial step in the analysis of the resulting data is the assignment of cells to gRNAs corresponding to a specific genetic intervention. However, this step is challenging due to a lack of systematic benchmarks and accessible software to apply and compare different guide assignment strategies. To address this, we here propose crispat (CRISPR guide assignment tool), a Python package to facilitate the choice of a suitable guide assignment strategy for single cell CRISPR screens. We demonstrate the package on four single cell CRISPR interference screens at low multiplicity of infection from two studies, where crispat identifies strong differences in the number of assigned cells, downregulation of the target genes and number of discoveries across different guide assignment strategies, highlighting the need for a suitable guide assignment strategy to obtain optimal power in single cell CRISPR screens. Availability and Implementation crispat is implemented in python and the source code, installation instructions and tutorials can be found at https://github.com/velten-group/crispat. Code to reproduce all findings in this paper is available at https://github.com/velten-group/crispat_analysis. Contact Britta Velten, britta.velten{at}cos.uni-heidelberg.de
Open AccessGuide assignment in single-cell CRISPR screens using crispat
Pooled single cell CRISPR screens have emerged as a powerful tool in functional genomics to probe the effect of genetic interventions at scale. A crucial step in the analysis of the resulting data is the assignment of cells to gRNAs corresponding to a specific genetic intervention. However, this step is challenging due to a lack of systematic benchmarks and accessible software to apply and compare different guide assignment strategies. To address this, we here propose crispat (CRISPR guide assignment tool), a Python package to facilitate the choice of a suitable guide assignment strategy for single cell CRISPR screens. We demonstrate the package on four single cell CRISPR interference screens at low multiplicity of infection from two studies, where crispat identifies strong differences in the number of assigned cells, downregulation of the target genes and number of discoveries across different guide assignment strategies, highlighting the need for a suitable guide assignment strategy to obtain optimal power in single cell CRISPR screens. Availability and Implementation crispat is implemented in python and the source code, installation instructions and tutorials can be found at https://github.com/velten-group/crispat. Code to reproduce all findings in this paper is available at https://github.com/velten-group/crispat_analysis. Contact Britta Velten, britta.velten{at}cos.uni-heidelberg.de