Biology subjects

Pfeifer, N.

Publications and source records attributed to Pfeifer, N..

13 recordsLinked to original sources

bioRxiv2026

Towards Useful and Private Synthetic Omics: Community Benchmarking of Generative Models for Transcriptomics Data

BackgroundThe synthesis of anonymized data derived from real-world cohorts offers a promising strategy for regulatory-compliant and privacy-preserving biological data sharing, potentially facilitating model development that can improve predictive performance. However, the extent to which generative models can preserve biological signals while remaining resilient to adversarial privacy attacks in high-dimensional omics contexts remains underexplored. To address this gap, the CAMDA 2025 Health Privacy Challenge launched a community-driven effort to systematically benchmark synthetic and privacy-preserving data generation for bulk RNA-seq cohorts. ResultsBuilding on this initiative, we systematically benchmarked 11 generative methods across two cancer cohorts ([~]1,000 and [~]5,000 patients) over 978 landmark genes. Methods were evaluated across complementary axes of distributional fidelity, downstream utility, biological plausibility and empirical privacy risk, with emphasis on trade-offs between vulnerability to membership inference attacks (MIA) and other evaluation dimensions. Expressive deep generative models achieved strong predictive utility and differential expression recovery, but were often more vulnerable to membership inference risk. Differentially private methods improved resistance to attacks at the cost of reduced utility, while simpler statistical approaches offered competitive utility with moderate privacy risk and fast training. ConclusionsSynthetic bulk RNA-seq quality is inherently multi-dimensional and shaped by trade-offs between utility, biological preservation and privacy. Our results indicate that differences in model architecture drive distinct trade-offs across these axes, suggesting that model choice should align with dataset characteristics, intended downstream use and privacy requirements. Privacy risk should also be assessed using multiple complementary attack methods and, where possible, formal differential privacy protection.

bioinformatics↗

bioRxiv2025

Predicting Drug Response with Multi-Task Gradient-Boosted Trees in Epilepsy

MotivationDespite the availability of numerous anti-seizure medications (ASMs), drug resistance remains a major issue for people with epilepsy. The probability of achieving seizure freedom diminishes with each unsuccessful drug trial, and the impact of genetic and clinical markers on ASM response remains unclear. To address this issue, we used state-of-the-art machine learning (ML) methods to predict the response of people with epilepsy to individual ASMs based on their clinical and genomic information. ResultsTo overcome data sparsity for less common drugs, we implement a multi-task (MT) learning approach for gradient-boosted trees (GBTs), assuming that predicting responses to different ASMs involves similar tasks. This strategy allows models for less prevalent drugs to leverage the more abundant data available for other drugs during training. The proposed model outperforms individual and combined drug-response predictions for most drugs. Our findings identify key genomic and clinical features influencing drug response, enhancing understanding of individual drug responses in people with epilepsy, and aiding clinicians in making informed treatment decisions. Availability and ImplementationDue to privacy reasons data is not publically available. The code will be made available upon acceptance under https://github.com/pfeiferAI/MT-GBT

bioinformatics↗

bioRxiv2022

Multi-view confounder detection for biomedical studies

In many biomedical studies an important first step is checking for confounding factors. For association studies, confounding can for example be caused by ethnic differences in the case and control groups. In many other settings there might be confounding factors like batch effects or founder effects that also need to be detected and controlled for1. Detecting confounding for data from one data source is well established (e.g., genomics data). Since more and more studies are now based on data from multiple data modalities (e.g., multi-omics), we evaluated whether multi-view confounder detection can benefit from state-of-the-art methods for multi-view data integration. Especially for clustering of multi-omics data, it has been shown that these methods can perform better than methods that treat the data modalities separately2. Our results show that multi-view confounder analysis is possible and that building on multi-view data integration methods is better than treating the different data modalities separately.

bioinformatics↗

bioRxiv2022

Learning with phenotypic similarity improves the prediction of functional effects of missense variants in voltage-gated sodium channels

BackgroundMissense variants in genes encoding voltage-gated sodium channels are associated with a spectrum of severe diseases affecting neuronal and muscle cells, the so-called sodium channelopathies. Variant effects on the biophysical function of the channel correlate with clinical features and can in most cases be categorized as an overall gain- or loss-of-function. This information enables a timely diagnosis, facilitates precision therapy, and guides prognosis. Machine learning models may be able to rapidly generate supporting evidence by predicting variant functional effects. MethodsHere, we describe a novel multi-task multi-kernel learning framework capable of harmonizing functional results and structural information with clinical phenotypes. We included 62 sequence- and structure-based features such as amino acid physiochemical properties, substitution radicality, conservation, protein-protein interaction sites, expert annotation, and others. We harmonized phenotypes as human phenotype ontology (HPO) terms, and compared different measures of phenotypic similarity under simulated sparsity or noise. The final model was trained on whole-cell patch-clamp recordings of 375 unique non-synonymous missense variants each expressed in mammalian cells. ResultsOur gain- or loss-of-function classifier outperformed both conventional baseline and state-of-the-art methods on internal validation (mean accuracy 0.837 {+/-} 0.035, mean AU-ROC 0.890 {+/-} 0.023) and on an independent set of recently described variants (n = 30, accuracy 0.967, AU-ROC 1.000). Model performance was robust across different phenotypic similarity measures and largely insensitive to phenotypic noise or sparsity. Localized multi-kernel learning offered biological insight and interpretability by highlighting channels with implicit genotype-phenotype correlations or latent task similarity for downstream analysis. ConclusionsLearning with phenotypic similarity makes efficient use of clinical information to enable accurate and robust prediction of variant functional effects. Our framework extends the use of human phenotype ontology terms towards kernel-based methods in machine learning. Training data, pre-trained models, and a web-based graphical user interface for the model are publicly available.

bioinformatics↗

bioRxiv2022

Insights to HIV-1 coreceptor usage by estimating HLA adaptation with Bayesian generalized linear mixed models

The mechanisms triggering the human immunodeficiency virus type I (HIV-1) to switch the coreceptor usage from CCR5 to CXCR4 during the course of infection are not entirely understood. While low CD4+ T cell counts are associated with CXCR4 usage, a predominance of CXCR4 usage with still high CD4+ T cell counts remains puzzling. Here, we explore the hypothesis that viral adaptation to the human leukocyte antigen (HLA) complex, especially to the HLA class II alleles, contributes to the coreceptor switch. To this end, we sequence the viral gag and env protein with corresponding HLA class I and II alleles of a new cohort of 312 treatment-naive, subtype C, chronically-infected HIV-1 patients from South Africa. To estimate HLA adaptation, we develop a novel computational approach using Bayesian generalized linear mixed models (GLMMs). Our model allows to consider the entire HLA repertoire without restricting the model to pre-learned HLA-polymorphisms as well as to correct for phylogenetic relatedness of the viruses within the model itself to account for founder effects. Using our model, we observe that CXCR4-using variants are more adapted than CCR5-using variants (p-value =1.34e-2). Additionally, adapted CCR5-using variants have a significantly lower predicted false positive rate (FPR) by the geno2pheno[coreceptor] tool compared to the non-adapted CCR5-using variants (p-value =2.21e-2), where a low FPR is associated with CXCR4 usage. Consequently, estimating HLA adaptation can be an asset in predicting not only coreceptor usage, but also an approaching coreceptor switch in CCR5-using variants. We propose the usage of Bayesian GLMMs for modeling virus-host adaptation in general. Author summaryViral control is currently our only counter mechanism against HIV-1 with no practicable cure nor a vaccine at hand. In treatment-naive patients, HLA adaptation and coreceptor usage of HIV-1 play a major role in their capability to control the virus. The interplay between both factors, however, has remained unexplored so far. Assessing the degree of viral HLA adaptation is challenging due to the exceptional genetic diversity of both the HLA complex and HIV-1. Therefore, current approaches constrain the adaptation prediction to a set of p-value selected HLA-polymorphism candidates. The selection of these candidates, however, requires extensive external large-scale population-based experiments that are not always available for the population of interest, especially not for newly emerging viruses. In this work, we present a novel computational approach using Bayesian generalized linear mixed models (GLMMs) that enables not only to predict the adaptation to the complete HLA profile of a patient, but also to handle phylogenetic-dependencies of the variants within the model directly. Using this light-weight approach for modeling (any) virus-host adaptation, we show that HLA adaptation is associated with coreceptor usage.

bioinformatics↗

bioRxiv2022

The characteristics of high-dark-diversity habitats derived from lidar

A key aspect of nature conservation is knowledge of which aspects of nature to conserve or restore to favor the characteristic diversity of plants in a given area. Here, we used a large plant dataset with > 40.000 plots combined with airborne laser scanning (lidar) data to reveal the local characteristics of habitats having a high plant dark diversity - i.e., absence of suitable species - at national extent (> 43.000 km2). Such habitats have potential for reaching high realized diversity levels and hence are important in a conservation context. We calculated 10 different lidar based metrics (both terrain and vegetation structure) and combined these with 7 different field-based measures (soil chemistry and species indicators). We then used Integrated Nested Laplace Approximation for modelling plant dark diversity across 33 North European habitat types (open landscapes and forests) selected by the European communities to be important. In open habitat types high-dark-diversity habitats had relatively low pH, high nitrogen content, tall homogenous vegetation and overall relatively homogenous terrains (high terrain openness) although with a rather high degree of local microtopographical variations. High-dark-diversity habitats in forests had relatively tall vegetation, few natural-forest indicators, low potential solar radiation input and a low cover of small woody plants. Our results highlight important vegetation, terrain and soil related factors that managers and policymakers should be aware of in conservation and restoration projects to ensure a natural plant diversity, for example low nutrient loads, natural microtopography and open forests with old-growth elements such as dead wood and rot attacks.

ecology↗

bioRxiv2021

Predicting the functional effects of voltage-gated potassium channel missense variants with multi-task learning

PurposeVariants in genes encoding voltage-gated potassium channels are associated with a broad spectrum of neurological diseases including epilepsy, ataxia, and intellectual disability. Knowledge of the resulting functional changes, characterized as overall ion channel gain- or loss-of-function, is essential to guide clinical management including precision medicine therapies. However, for an increasing number of variants, little to no experimental data is available. New tools are needed to evaluate variant functional effects. MethodsWe catalogued a comprehensive dataset of 959 functional experiments across 19 voltage-gated potassium channels, leveraging data from 782 unique disease-associated and synthetic variants. We used these data to train a taxonomy-based multi-task learning support vector machine (MTL-SVM), and compared performance to a baseline of standard SVMs. ResultsMTL-SVM maintains channel family structure during model training, improving overall predictive performance (mean balanced accuracy 0.729 {+/-} 0.029, AU-ROC 0.757 {+/-} 0.039) over baseline (mean balanced accuracy 0.645 {+/-} 0.041, AU-ROC 0.710 {+/-} 0.074). We can obtain meaningful predictions even for channels with few known variants (KCNC1, KCNQ5). ConclusionOur model enables functional variant prediction for voltage-gated potassium channels. It may assist in tailoring current and future precision therapies for the increasing number of patients with ion channel disorders.

bioinformatics↗

bioRxiv2021

Kinetics and correlates of the neutralizing antibody response to SARS-CoV-2

A detailed understanding of antibody-based SARS-CoV-2 immunity has critical implications for overcoming the COVID-19 pandemic and for informing on vaccination strategies. In this study, we evaluated the dynamics of the SARS-CoV-2 antibody response in a cohort of 963 recovered individuals over a period of 10 months. Investigating a total of 2,146 samples, we detected an initial SARS-CoV-2 antibody response in 94.4% of individuals, with 82% and 79% exhibiting serum and IgG neutralization, respectively. Approximately 3% of recovered patients demonstrated exceptional SARS-CoV-2 neutralizing activity, defining them as elite neutralizers. These individuals also possessed effective cross-neutralizing IgG antibodies to SARS-CoV-1 without any known prior exposure to this virus. By applying multivariate statistical modeling, we found that sero-reactivity, age, time since disease onset, and fever are key factors predicting SARS-CoV-2 neutralizing activity in mild courses of COVID-19. Investigating longevity of the antibody response, we detected loss of anti-spike reactivity in 13% of individuals 10 months after infection. Moreover, neutralizing activity had an initial half-life of 6.7 weeks in serum versus 30.8 weeks in purified IgG samples indicating the presence of a more stable and long-term memory IgG B cell repertoire in the majority of individuals recovered from COVID-19. Our results demonstrate a broad spectrum of the initial SARS-CoV-2 neutralizing antibody response depending on clinical characteristics, with antibodies being maintained in the majority of individuals for the first 10 months after mild course of COVID-19.

immunology↗

bioRxiv2020

Investigating the drivers of macro-fungal dark diversity using LiDAR

Despite the important role of fungi for ecosystems, relatively little is known about the factors underlying the dynamics of their diversity. Moreover, studies do not typically consider their dark diversity: the species absent from an otherwise suitable site. Here, we examined potential drivers of local fungal dark diversity in temperate woodland and open habitats using LiDAR and in-situ field measurements, combined with a systematically collected and geographically comprehensive macro-fungi and plant data set. For the first time, we also estimated species pools of fungi by considering both plant and fungi co-occurrences. The most important LiDAR variables for modelling fungal dark diversity were amplitude and echo ratio, which are both thought to represent vegetation structure. These results suggest that the local fungal dark diversity is highest in production forests like plantations and lowest in more open forests and in open habitats with little woody vegetation. Plant species richness was the strongest explanatory factor overall and negatively correlated with local fungal dark diversity. Soil fertility showed a positive relationship with dark diversity in open habitats. These findings may indicate that the local dark diversity of macro-fungi is highest in areas with a relatively high human impact (typically areas with low plant species richness and high soil fertility). Overall, this study brings novel insights into local macro-fungi dark diversity patterns, suggesting that a multitude of drivers related to both soil and vegetation act in concert to determine fungal dark diversity.

ecology↗

bioRxiv2020

Longitudinal isolation of potent near-germline SARS-CoV-2-neutralizing antibodies from COVID-19 patients

The SARS-CoV-2 pandemic has unprecedented implications for public health, social life, and world economy. Since approved drugs and vaccines are not available, new options for COVID-19 treatment and prevention are highly demanded. To identify SARS-CoV-2 neutralizing antibodies, we analysed the antibody response of 12 COVID-19 patients from 8 to 69 days post diagnosis. By screening 4,313 SARS-CoV-2-reactive B cells, we isolated 255 antibodies from different time points as early as 8 days post diagnosis. Among these, 28 potently neutralized authentic SARS-CoV-2 (IC100 as low as 0.04 g/ml), showing a broad spectrum of V genes and low levels of somatic mutations. Interestingly, potential precursors were identified in naive B cell repertoires from 48 healthy individuals that were sampled before the COVID-19 pandemic. Our results demonstrate that SARS-CoV-2 neutralizing antibodies are readily generated from a diverse pool of precursors, fostering the hope of rapid induction of a protective immune response upon vaccination.

immunology↗

bioRxiv2019

openPrimeR for multiplex amplification of highly diverse templates

To study the diversity of immune receptors and pathogens, multiplex PCR has become a central approach in research and diagnostics. However, insufficient primer design against highly diverse templates often prevents amplification and therefore limits the correct understanding of biological processes. Here, we present openPrimeR, an R-based tool for evaluating and designing multiplex PCR primers. openPrimeR provides a functional and intuitive interface and uses either a greedy algorithm or an integer linear program to compute the minimal set of primers that performs full target coverage. As proof of concept, we used openPrimeR to find optimal primer sets for the amplification of highly mutated immunoglobulins. Comprehensive analyses on specifically generated immunoglobulin variable gene segment libraries resulted in the composition of highly effective primer sets (oPR-IGHV, oPR-IGKV and oPR-IGLV) that demonstrated to be particularly suitable for the isolation of novel human antibodies.

immunology↗

bioRxiv2019

Unique and assay specific features of NOMe-, ATAC- and DNase I-seq data

Chromatin accessibility maps are important for the functional interpretation of the genome. Here, we systematically analysed assay specific differences between DNase I-Seq, ATAC-Seq and NOMe-Seq in a side by side experimental and bioinformatic setup. We observe that most prominent nucleosome depleted regions (NDRs, e.g. in promoters) are roboustly called by all three or at least two assays. However we also find a high proportion of assay specific NDRs that are often "called" by only one of the assays. We show evidence that these assay specific NDRs are indeed genuine open chromatin sites and contribute important information for accurate gene expression prediction. While technically ATAC-Seq and DNAse I-Seq provide a high NDR calling rate for relatively low sequencing costs in comparison to NOMe-Seq, NOMe-Seq singles out as it provides a multitude of information: it allows to not only detect NDRs but also endogenous DNA methylation, genome wide segmentation into heterochromatic A/B domains and local phasing of nucleosomes outside of NDRs. In summary our comparison strongly suggest to consider assay specific differences for the experimental desgin and for generalized and comparative functional interpretations.

genomics↗

bioRxiv2017

All Fingers Are Not The Same: Handling Variable-Length Sequences In A Discriminative Setting Using Conformal Multi-Instance Kernels

Most string kernels for comparison of genomic sequences are generally tied to using (absolute) positional information of the features in the individual sequences. This poses limitations when comparing variable-length sequences using such string kernels. For example, profiling chromatin interactions by 3C-based experiments results in variable-length genomic sequences (restriction fragments). Here, exact position-wise occurrence of signals in sequences may not be as important as in the scenario of analysis of the promoter sequences, that typically have a transcription start site as reference. Existing position-aware string kernels have been shown to be useful for the latter scenario.\n\nIn this work, we propose a novel approach for sequence comparison that enables larger positional freedom than most of the existing approaches, can identify a possibly dispersed set of features in comparing variable-length sequences, and can handle both the aforementioned scenarios. Our approach, CoMIK, identifies not just the features useful towards classification but also their locations in the variable-length sequences, as evidenced by the results of three binary classification experiments, aided by recently introduced visualization techniques. Furthermore, we show that we are able to efficiently retrieve and interpret the weight vector for the complex setting of multiple multi-instance kernels.

bioinformatics↗