Biology subjects

Beerenwinkel, N.

Publications and source records attributed to Beerenwinkel, N..

At least 19 recordsLinked to original sources

bioRxiv2026

Reliable single-cell perturbations explain and improve model performance

Predicting single-cell transcriptional responses to perturbations is central to building the virtual cell, yet recent benchmarks show that simple baseline methods often outperform complex models, and model comparisons depend on the evaluation metric. Most studies assume that preprocessed RNA sequencing data are reliable ground truth for both training and evaluation. Here, we test this assumption by measuring the reliability of perturbations and their alignment with shared perturbation responses, classifying each perturbation as specific, shared, or unreliable. Among 7,170 perturbations from 29 datasets, 65% are unreliable, 11% shared, and 24% specific. Applying these quality labels to published benchmarks shows that model comparisons depend on perturbation quality. Training with reliable perturbations alone matches or outperforms full-data performance while using 55% of all training perturbations. Our framework also enables prospective experimental design: for most perturbations, a 28-cell pilot experiment accurately predicts how many cells a full screen needs to be reliable.

bioinformatics↗

bioRxiv2026

Clone-level multi-modal prediction of tumour drug response

Tumour heterogeneity presents a major challenge for precision oncology, as genetically and phenotypically distinct tumour clones may respond differently to therapy. To address this, we introduce scClone2DR, a probabilistic multi-modal framework that predicts drug responses at the level of individual tumour clones by integrating single-cell DNA and RNA sequencing with ex-vivo drug-screening data. In simulations, scClone2DR substantially outperforms alternatives in recovering true drug-effects and clonal sensitivities. Applied to 60 melanoma and 21 acute myeloid leukaemia patient samples, the method identifies heterogeneous clonal responses, yields biologically meaningful feature rankings, highlights clones that may be resistant to treatment, and improves the prediction of clinical outcomes compared to models ignoring clonal structure. These results demonstrate that modelling tumour evolution and clonal diversity is crucial for accurate drug-response prediction and provides a foundation for more effective, clone-aware precision oncology.

cancer biology↗

bioRxiv2026

Domain-adaptation deep learning models do not outperform simple baseline models in single-cell anti-cancer drug sensitivity prediction

Tumor drug response is profoundly shaped by cellular heterogeneity, making single-cell resolution essential for precision oncology. While drug-response labels are abundant for cell lines at bulk resolution, translating these predictive models to the single-cell level requires effective domain adaptation strategies. Motivated by advances in computer vision, recent deep-learning domain adaptation methods promise to transfer knowledge from bulk (source) to single-cell (target) data with-out the need for target labels. However, their true translational utility remains unclear due to a lack of rigorous evaluation against non-adaptive baselines across diverse biological and technical contexts. Here, we present a comprehensive benchmark comparing four representative domain adaptation methods against two simple gradient boosting baseline methods. Through systematic evaluation across 19 single-cell datasets and 10 drugs, we show that none of the complex adaptation methods outperforms the simpler baselines. By analyzing the drivers of model performance, we find that target-informed hyperparameter tuning and sparse label supervision are the principal sources of prediction gain. Our study reveals that current approaches fail to bridge the bulk-to-single-cell conceptual shift and provides a unified codebase and comprehensive data collection to facilitate robust model comparisons. By enabling transparent evaluation and robust benchmarking against simple models, this resource aims to accelerate future developments in translational pharmacogenomics.

cancer biology↗

bioRxiv2025

Phylogenetic tree inference from single-cell RNA sequencing data

Single-cell RNA sequencing technologies enable the large-scale measurement of gene expression profiles at the individual cell level to assess cellular diversity and function. In oncology, leveraging these single-cell transcriptomic data to reconstruct the phylogenetic relationships among cancer cells can provide insights into tumor evolution, metastasis formation, and the development of treatment resistance. However, phylogenetic inference from single-cell RNA sequencing is challenging due to sparse and noisy data and large dataset sizes. We present a novel tree inference method designed for such data that takes reference and alternative read counts of single-nucleotide variants and reconstructs a phylogenetic tree of the sequenced cells via maximum likelihood using a random-scan greedy search. To overcome local optima in the search, our algorithm alternates between two different tree representations: cell lineage trees, where cells are represented by nodes and mutations are attached to edges, and mutation trees, where mutation nodes encode the mutational events and cells are attached to them. Because a local optimum in one tree space generally does not correspond to a local optimum in the other space, we maximize the likelihood by switching between the two tree spaces until convergence is achieved in both. We demonstrate superior performance of our approach on simulated data with complex clonal architectures compared to existing methods. Furthermore, we show its applicability to cancer single-cell RNA sequencing data, which allows us to link evolutionary trajectories of cells to their gene expression profiles.

genomics↗

bioRxiv2025

Assessing genomic reproducibility of read alignment tools

Genomic research relies on accurate and reproducible computational analyses of DNA sequencing data to draw reliable biological conclusions. Read mapping, the process of aligning reads to a reference genome, is central to many applications, including variant detection and comparative genomics. While several tools have been developed for this task, genomic reproducibility1, defined as the consistency of results across replicates, remains underexplored. Here, we address this question by introducing a methodology based on synthetic replicates of sequencing data, generated by perturbing the original reads through shuffling, reverse complementing, or combined shuffling and reverse complementing. Our approach is able to simulate variability observed across sequencing runs due to differences in library preparation techniques. We evaluated the reproducibility of eight alignment tools (BWA-MEM22, Bowtie23, HISAT24, minimap25, NextGenMap6, SNAP7, SMALT7,8 and Subread9) under these perturbations using whole-genome sequencing (WGS) data. Synthetic replicates were aligned and compared to the original sample to quantify discrepancies. Mapping accuracy changes ranged from 0.0001% to 4.4% for primary reads, which are alignments not marked as secondary, supplementary, or duplicates, and up to 12.2% for high-quality primary reads. For primary reads, the percentage of reads commonly mapped in both the original and synthetic replicate ranged from 91.66% to 100% relative to the total number of mapped reads in the original dataset. High-quality filtering improved consistency, though some tools still failed to recover more than 70% of the original alignments. Within the set of common reads, the incidence of inconsistent mappings was as high as 13.53% for primary reads and 6.73% for high-quality primary reads. Bowtie 2 was fully reproducible under the shuffling replicate, Subread was fully reproducible under the reverse-complement replicate, whereas NextGenMap exhibited only minor inconsistencies. By contrast, SNAP and minimap2 displayed the most significant variability under reverse complementing. We further demonstrate that alignment inconsistencies propagate to the downstream task of calling structural genomic variants. Using Manta for structural-variant (SV) calling, we observed that Bowtie 2, HISAT2, and minimap2 maintained perfect SV concordance between original and replicate alignments, whereas other tools exhibited lower concordance, with Subread scoring only 87% concordance. In conclusion, our comprehensive evaluation demonstrates that synthetic perturbations reveal critical differences in how alignment tools handle technical variability and how these differences propagate to downstream variant analyses, underscoring the necessity of incorporating reproducibility benchmarks into the selection and validation of read mappers to ensure robust and reliable genomic interpretations.

bioinformatics↗

bioRxiv2025

Cell lineage inference from mitochondrial mutations

Tracing genealogical relationships between cells and their molecular states is essential for understanding tissue development, maintenance and disease. We present mt-SCITE, a computational method that reconstructs cell division histories from mitochondrial mutations in single cells. On simulated and experimental benchmarks, mt-SCITE outperformed existing methods in terms of cell lineage reconstruction, and scaled to thousands of cells. Applied to expanded CD8+ T cells following vaccination, mt-SCITE resolved subclonal relationships and revealed clones with predominantly memory-like, effector-like, or mixed states. mt-SCITE robustly identifies clonally related populations and resolves substructure within them, providing a broadly applicable approach to study cell lineage development.

bioinformatics↗

bioRxiv2025

Bayesian inference of fitness landscapes via tree-structured branching processes

MotivationThe complex dynamics of cancer evolution, driven by mutation and selection, underlies the molecular heterogeneity observed in tumors. The evolutionary histories of tumors of different patients can be encoded as mutation trees and reconstructed in high resolution from single-cell sequencing data, offering crucial insights for studying fitness effects of and epistasis among mutations. Existing models, however, either fail to separate mutation and selection or neglect the evolutionary histories encoded by the tumor phylogenetic trees. ResultsWe introduce FiTree, a tree-structured multi-type branching process model with epistatic fitness parameterization and a Bayesian inference scheme to learn fitness landscapes from single-cell tumor mutation trees. Through simulations, we demonstrate that FiTree outperforms state-of-the-art methods in inferring the fitness landscape underlying tumor evolution. Applying FiTree to a single-cell acute myeloid leukemia dataset, we identify epistatic fitness effects consistent with known biological findings and quantify uncertainty in predicting future mutational events. The new model unifies probabilistic graphical models of cancer progression with population genetics, offering a principled framework for understanding tumor evolution and informing therapeutic strategies. Availability and implementationThe Python package FiTree and the analysis workflows are available at https://github.com/cbg-ethz/FiTree.

bioinformatics↗

bioRxiv2024

DemoTape: Computational demultiplexing of targeted single-cell sequencing data

BackgroundSingle-cell sequencing can provide novel insights into the understanding and treatment of diseases. In cancer, for example, intratumor heterogeneity is a major cause of treatment resistance and relapse. Although technological progress has substantially increased the throughput of sequenced cells, single-cell sequencing remains cost and labor-intensive. Multiplexing, i.e., the pooling and subsequent joint preparation and sequencing of samples, followed by a demultiplexing step, is a common practice to reduce expenses and confounding batch effects, especially in single-cell RNA sequencing. ResultsHere, we introduce demoTape, a computational demultiplexing method for targeted single-cell DNA sequencing (scDNA-seq) data based on a distance metric between individual cells at single-nucleotide polymorphisms loci. To validate demoTape, we sequence three B-cell lymphoma patients separately and multiplexed on the Tapestri platform. We find similar genotypes, clones, and evolutionary histories in all three samples when comparing the individual with the demultiplexed samples. Using the three individually sequenced samples, we simulate multiplexed ground truth data and show that demoTape outperforms state-of-the-art demultiplexing methods designed for RNA sequencing data. Additionally, we demonstrate through downsampling that the inferred clonal composition remained largely stable for samples with fewer cells despite the inevitable loss in resolution of low-frequency clones. ConclusionsMultiplexing and subsequent genotype-based demultiplexing of scDNA-seq will reduce costs and workload, eventually allowing the sequencing of more samples. This will open new possibilities and accelerate the investigation of biological questions where cellular heterogeneity on the genomic level plays a crucial role.

bioinformatics↗

bioRxiv2024

Robust and cost-efficient single-cell sequencing through combinatorial pooling

Single-cell sequencing is widely used to study molecular cell-to-cell heterogeneity. Even though the cost of sequencing has dropped throughout the last decades, single-cell assays remain expensive, because they require strategies to index molecules by cells. The high costs of indexing can be mitigated by pooling samples prior to sequencing library preparation. Computational methods have been developed to leverage molecular features that are distinct between different samples to separate the pools into distinct datasets. However, since all multiplexed samples are processed in the same way, information on the origin of each demultiplexed dataset is lost. To map datasets to their sample of origin, additional information such as molecular indexing or additional genotyping is needed. Here, we propose a class of experimental designs that allows identifying the sample of origin of each demultiplexed dataset, only relying on the genetic profiles of the samples and the composition of pools. Our approach is based on splitting and pooling samples in specific combinations. We find a most cost-efficient experimental design in this class and prove its optimality. We present a dynamic programming algorithm to iteratively simplify an optimal experimental design by breaking it into several independent designs while maintaining optimality. Furthermore, we propose a subclass of experimental designs which allow robust sample identification even under partial failure of the experiment and present a provably optimal design in this subclass. We provide an implementation for automatic sample identification under these optimal combinatorial pooling strategies and demonstrate its functionality in a simulation study.

bioinformatics↗

bioRxiv2024

OncotreeVIS - an interactive graphical user interface for visualizing mutation tree cohorts

MotivationIn recent years, developments in single-cell next-generation sequencing technology and computational methodology have made it possible to reconstruct, with increasing precision, the evolutionary history of tumors and their cell phylogenies, represented as mutation trees. Many mutation tree inference tools exist, but they do not support detailed visual tree inspection, nor tree comparisons or analysis at the cohort level, an important task in computational oncology. ResultsWe developed oncotreeVIS, an interactive graphical user interface for visualizing mutation tree cohorts and tree posterior distributions obtained from mutation tree inference tools. Oncotree-VIS can display mutation trees that encode single or joint genetic events, such as point mutations and copy number changes, and highlight matching subclones, conserved trajectories and drug-gene interactions at cohort level. OncotreeVIS facilitates the visual inspection of mutation tree clusters and pairwise tree distances. It is available both as a JavaScript library that can be used locally or as a web application that can be accessed online. It includes seven default datasets of public mutation tree cohorts for visualization, while new mutation trees are provided in a predefined JSON format. Availability and implementationhttps://github.com/cbg-ethz/oncotreeVIS. Web applicationhttps://cbg-ethz.github.io/oncotreeVIS.

cancer biology↗

bioRxiv2024

Bayesian modeling of mutual exclusivity in cancer mutations

When cancer develops, gene mutations do not occur independently, prompting re-searchers to pose scientific hypotheses about their interactions. Synthetic lethal interactions, which result in mutually exclusive mutations, have received considerable attention as they may inform about the structure of aberrant biological pathways in cancer cells and suggest therapeutic targets. However, finding patterns of mutually exclusive genes is a challenging task due to small available sample sizes, sequencing noise, and confounders present in observational studies. Here, we leverage recent advancements in probabilistic programming to propose a fully Bayesian framework for modeling mutual exclusivity based on a family of constrained Bernoulli mixture models. By forming continuous model expansion within the iterative Bayesian workflow, we quantify the uncertainty resulting from small sample sizes and perform careful model criticism. Our analysis indicates that alterations in the EGFR and IDH1 genes may exhibit mutual exclusivity in glioblastoma multiforme tumors. We argue that Bayesian analysis offers a conceptual, systematic, and computationally feasible approach to model building, complementing the findings obtained from classical hypothesis testing approaches. Code: https://github.com/cbg-ethz/jnotype

bioinformatics↗

bioRxiv2024

VILOCA: Sequencing quality-aware haplotype reconstruction and mutation calling for short- and long-read data

RNA viruses exist in large heterogeneous populations within their host. The structure and diversity of virus populations affects disease progression and treatment outcomes. Next-generation sequencing allows detailed viral population analysis, but inferring diversity from error-prone reads is challenging. Here, we present VILOCA, a method for mutation calling and reconstruction of local haplotypes from short- and long-read viral sequencing data. Local haplotypes refer to genomic regions that have approximately the length of the input reads. VILOCA recovers local haplotypes by using a Dirichlet process mixture model to cluster reads around their unobserved haplotypes and leveraging quality scores of the sequencing reads. We assessed the performance of VILOCA in terms of mutation calling and haplotype reconstruction accuracy on simulated and experimental Illumina, PacBio, and Oxford Nanopore data. On simulated and experimental Illumina data, VILOCA performed better or similar to existing methods. On the simulated long-read data, VILOCA is able to recover on average 82% of the ground truth mutations with perfect precision compared to only 64% recall and 90% precision of the second-best method. In summary, VILOCA provides significantly improved accuracy in mutation and haplotype calling, especially for long-read sequencing data, and therefore facilitates the comprehensive characterization of heterogeneous within-host viral populations.

bioinformatics↗

bioRxiv2024

De novo detection of somatic variants in long-read single-cell RNA sequencing data

In cancer, genetic and transcriptomic variations generate clonal heterogeneity, leading to treatment resistance. Long-read single-cell RNA sequencing (LR scRNA-seq) has the potential to detect genetic and transcriptomic variations simultaneously. Here, we present LongSom, a computational workflow leveraging high-quality LR scRNA-seq data to call de novo somatic single-nucleotide variants (SNVs), including in mitochondria (mtSNVs), copy-number alterations (CNAs), and gene fusions, to reconstruct the tumor clonal heterogeneity. Before somatic variants calling, LongSom re-annotates marker gene based cell types using cell mutational profiles. LongSom distinguishes somatic SNVs from noise and germline polymorphisms by applying an extensive set of hard filters and statistical tests. Applying LongSom to human ovarian cancer samples, we detected clinically relevant somatic SNVs that were validated against matched DNA samples. Leveraging somatic SNVs and fusions, LongSom found subclones with different predicted treatment outcomes. In summary, LongSom enables de novo variants detection without the need for normal samples, facilitating the study of cancer evolution, clonal heterogeneity, and treatment resistance.

cancer biology↗

bioRxiv2024

CTAT-LR-fusion: accurate fusion transcript identification from long and short read isoform sequencing at bulk or single cell resolution

Gene fusions are found as cancer drivers in diverse adult and pediatric cancers. Accurate detection of fusion transcripts is essential in cancer clinical diagnostics, prognostics, and for guiding therapeutic development. Most currently available methods for fusion transcript detection are compatible with Illumina RNA-seq involving highly accurate short read sequences. Recent advances in long read isoform sequencing enable the detection of fusion transcripts at unprecedented resolution in bulk and single cell samples. Here we developed a new computational tool CTAT-LR-fusion to detect fusion transcripts from long read RNA-seq with or without companion short reads, with applications to bulk or single cell transcriptomes. We demonstrate that CTAT-LR-fusion exceeds fusion detection accuracy of alternative methods as benchmarked with simulated and real long read RNA-seq. Using short and long read RNA-seq, we further apply CTAT-LR-fusion to bulk transcriptomes of nine tumor cell lines, and to tumor single cells derived from a melanoma sample and three metastatic high grade serous ovarian carcinoma samples. In both bulk and in single cell RNA-seq, long isoform reads yielded higher sensitivity for fusion detection than short reads with notable exceptions. By combining short and long reads in CTAT-LR-fusion, we are able to further maximize detection of fusion splicing isoforms and fusion-expressing tumor cells. CTAT-LR-fusion is available at https://github.com/TrinityCTAT/CTAT-LR-fusion/wiki.

cancer biology↗

bioRxiv2024

Modeling metastatic progression from cross-sectional cancer genomics data

Metastasis formation is a hallmark of cancer lethality. Yet, metastases are generally unobservable during their early stages of dissemination and spread to distant organs. Genomic datasets of matched primary tumors and metastases may offer insights into the underpinnings and the dynamics of metastasis formation. We present metMHN, a cancer progression model designed to deduce the joint progression of primary tumors and metastases using cross-sectional cancer genomics data. The model elucidates the statistical dependencies among genomic events, the formation of metastasis, and the clinical emergence of both primary tumors and their metastatic counterparts. metMHN enables the chronological reconstruction of mutational sequences and facilitates estimation of the timing of metastatic seeding. In a study of nearly 5000 lung adenocarcinomas, metMHN pinpointed TP53 and EGFR as mediators of metastasis formation. Furthermore, the study revealed that post-seeding adaptation is predominantly influenced by frequent copy number alterations. All datasets and code are available on GitHub at https://github.com/cbg-ethz/metMHN.

bioinformatics↗

bioRxiv2023

Tracking SARS-CoV-2 variants of concern in wastewater: an assessment of nine computational tools using simulated genomic data

Wastewater-based surveillance (WBS) is an important epidemiological and public health tool for tracking pathogens across the scale of a building, neighbourhood, city, or region. WBS gained widespread adoption globally during the SARS-CoV-2 pandemic for estimating community infection levels by qPCR. Sequencing pathogen genes or genomes from wastewater adds information about pathogen genetic diversity which can be used to identify viral lineages (including variants of concern) that are circulating in a local population. Capturing the genetic diversity by WBS sequencing is not trivial, as wastewater samples often contain a diverse mixture of viral lineages with real mutations and sequencing errors, which must be deconvoluted computationally from short sequencing reads. In this study we assess nine different computational tools that have recently been developed to address this challenge. We simulated 100 wastewater sequence samples consisting of SARS-CoV-2 BA.1, BA.2, and Delta lineages, in various mixtures, as well as a Delta-Omicron recombinant and a synthetic "novel" lineage. Most tools performed well in identifying the true lineages present and estimating their relative abundances, and were generally robust to variation in sequencing depth and read length. While many tools identified lineages present down to 1% frequency, results were more reliable above a 5% threshold. The presence of an unknown synthetic lineage, which represents an unclassified SARS-CoV-2 lineage, increases the error in relative abundance estimates of other lineages, but the magnitude of this effect was small for most tools. The tools also varied in how they labelled novel synthetic lineages and recombinants. While our simulated dataset represents just one of many possible use cases for these methods, we hope it helps users understand potential sources of noise or bias in wastewater sequencing data and to appreciate the commonalities and differences across methods.

bioinformatics↗

bioRxiv2023

Overcoming Observation Bias for Cancer Progression Modeling

Cancers evolve by accumulating genetic alterations, such as mutations and copy number changes. The chronological order of these events is important for understanding the disease, but not directly observable from cross-sectional genomic data. Cancer progression models (CPMs), such as Mutual Hazard Networks (MHNs), reconstruct the progression dynamics of tumors by learning a network of causal interactions between genetic events from their co-occurrence patterns. However, current CPMs fail to include effects of genetic events on the observation of the tumor itself and assume that observation occurs independently of all genetic events. Since a dataset contains by definition only tumors at their moment of observation, neglecting any causal effects on this event leads to the "conditioning on a collider" bias: Events that make the tumor more likely to be observed appear anti-correlated, which results in spurious suppressive effects or masks promoting effects among genetic events. Here, we extend MHNs by modeling effects from genetic progression events on the observation event, thereby correcting for the collider bias. We derive an efficient tensor formula for the likelihood function and learn two models on somatic mutation datasets from the MSK-IMPACT study. In colon adenocarcinoma, we find a strong effect on observation by mutations in TP53, and in lung adenocarcinoma by mutations in EGFR. Compared to classical MHNs, this explains away many spurious suppressive interactions and uncovers several promoting effects. The data, code, and results are available at https://github.com/cbg-ethz/ObservationMHN.

bioinformatics↗

bioRxiv2023

oncotree2vec - A method for embedding and clustering of tumor phylogenetic trees

Understanding the genomic heterogeneity of tumors is an important task in computational oncology, especially in the context of finding personalized treatments based on the genetic profile of each patients tumor. Tumor clustering that takes into account the temporal order of genetic events, as represented by tumor mutation trees, is a powerful approach for grouping together patients with genetically and evolutionarily similar tumors and can provide insights into discovering tumor sub-types, for more accurate clinical diagnosis and prognosis. Here, we propose oncotree2vec, a method for clustering tumor mutation trees by learning vector representations of mutation trees that capture the different relationships between subclones in an unsupervised manner. Learning low-dimensional tree embeddings facilitates the visualization of relations between trees in large cohorts and can be used for downstream analyses, such as deep learning approaches for single-cell multi-omics data integration. We assessed the performance and the usefulness of our method in three simulation studies, and on two real datasets: a cohort of 43 trees from six cancer types with different branching patterns corresponding to different modes of spatial tumor evolution and a cohort of 123 AML mutation trees.

bioinformatics↗