Biology subjects

Baumbach, J.

Publications and source records attributed to Baumbach, J..

At least 19 recordsLinked to original sources

bioRxiv2026

Concurrent de novo inference of cell types and gene regulatory networks with NetMap

Existing methods for GRN inference based on single-cell RNA-sequencing (scRNA-seq) data typically require prior cell type calling and then infer GRNs at cell type resolution. This risks propagation of errors from cell annotation to GRN inference. Moreover, it makes existing methods blind to regulatory programs that are shared between different cell types, as well as to subtle differences in regulatory activity between cells of the same type. To address these limitations, we present NetMap, a pipeline that jointly infers cell-level GRNs and clusters cells based on shared regulatory programs, combining state-of-the-art single-cell autoencoders and explainable artificial intelligence methods. NetMap outperforms existing methods on synthetic data with a known ground-truth GRN and is able to reliably detect biologically plausible regulons on real data. Moreover, requiring only a small set of user-provided marker genes, NetMap can be used for GRN structure-informed probabilistic cell type assignment. We demonstrate the utility of this approach using PBMC data, where we achieve near perfect clustering and cell type recovery performance compared with manually created cell type labels. Finally, we applied NetMap to a data set with naive, activated and exhausted T-cells, and were able to verify that NetMap retrieves relevant T-cell state signatures from the data, with two approaches. NetMap can be installed as a python package from https://github.com/bionetslab/netmap and is compatible with the scverse system. Scripts to reproduce our results can be found at https://github.com/bionetslab/netmap-evaluation.

bioinformatics↗

bioRxiv2026

CyChat: a conversational Cytoscape app for no-code, reproducible network analysis

Network-based analyses of molecular interactions are useful for interpreting high-throughput omics data and identifying therapeutic targets. Cytoscape is the standard platform for these tasks, but users face a trade-off between accessible graphical workflows that are difficult to document and reproducible automation in Python or R that requires programming expertise. General-purpose coding assistants can generate Cytoscape Automation scripts, but remain external to Cytoscape. We present CyChat, a Cytoscape Desktop app that integrates a chat interface and a large language model (LLM) agent into the application. CyChat translates natural language into executable Cytoscape Automation workflows, runs generated Python code, and exports chat sessions with executed code as standalone Jupyter notebooks. To reduce setup barriers, CyChat includes an embedded Python runtime and supports both cloud-based and locally hosted LLMs. CyChat was evaluated across ten Cytoscape workflows using seven LLM providers, each represented by one LLM. The strongest configuration achieves a pass rate above 99%. In a qualitative evaluation based on a published network visualization, CyChat completes the task in 1.5-5 minutes, compared with 15-20 minutes for manual GUI workflows by computational biologists. CyChat is available through the Cytoscape App Store at https://apps.cytoscape.org/apps/cychat.

bioinformatics↗

bioRxiv2026

Unraveling the Comedone Switch through Single-Cell Resolution of Human Acne Lesions

Acne vulgaris is one of the most prevalent inflammatory skin diseases worldwide, yet the molecular events initiating comedogenesis remain poorly understood. The comedone switch hypothesis proposes that acne originates from an imbalance in lineage commitment within the junctional zone of the pilosebaceous unit, promoting infundibular differentiation at the expense of sebaceous gland maintenance. However, direct evidence from human acne tissue at single-cell resolution has been lacking. Here, we integrated single-cell transcriptomic datasets from healthy skin, non-lesional skin of acne patients, and lesional acne tissue to reconstruct the earliest stages of comedogenesis. We identified a previously uncharacterized cell population in non-lesional skin with transcriptomic features consistent with a microcomedone and mapped this population across independent datasets to reconstruct the transcriptional comedone architecture. Comedonal remodeling was characterized by enhanced keratinization and inflammatory programs. Quantitative analyses supported a shift from sebaceous toward infundibular cell fate, providing first data-driven evidence for the comedone switch hypothesis in human acne. Beyond the pilosebaceous unit, we identified broader epithelial alterations, including loss of POSTN and ERRFI1 expression in basal interfollicular epidermal keratinocytes. Together, these findings provide a cell-resolved framework for human comedogenesis and identify candidate mechanisms linking genetic susceptibility, environmental triggers, and lineage imbalance within the upper hair follicle.

bioinformatics↗

bioRxiv2026

Automated landmark and semilandmark annotation for wing geometric morphometrics in Diptera using deep learning

1. Diptera represent a diverse insect order, including vectors of human and animal pathogens. Their accurate species identification remains a major bottleneck in ecological and epidemiological studies. Morphological identification requires taxonomic expertise, while molecular methods are costly and not universally reliable. Wing geometric morphometrics offers an alternative, but manual landmark annotation is time-consuming and introduces observer bias. 2. We developed ITHILDIN, an automated pipeline for landmark and semilandmark annotation of Diptera wings, combining UNet++ segmentation and an Hourglass landmark prediction model. Using mosquitoes as the primary model system, we extended an existing repository with 5,793 additional images. Models were trained on 5991 annotations of landmarks and segmentations and then evaluated on 12,522 images across 34 taxa. We assessed landmark prediction accuracy against human observers and ML-morph, evaluated species identification using Linear Discriminant Analysis on 17 homologous landmarks and 52 semilandmarks, and tested out-of-distribution generalisation by reproducing an independent study. Transferability was demonstrated by adapting the pipeline to the Dipteran families Drosophilidae and Glossinidae. 3. The Hourglass model achieved a mean landmark error of 4.5 pixels (95% CI: 4.3-4.6), within human observer variability (4.7 pixels, 95% CI: 4.4-5.0) and substantially outperforming ML-Morph (12.7 pixels, 95% CI: 11.1-14.2). The semilandmark-based approach for species identification achieved 91% balanced accuracy across 34 taxa, comparable to CNN performance (94%). On out-of-distribution data, the landmark pipeline generalised substantially better than the CNN and a soft-voting ensemble of the landmark and CNN classifiers achieved 88% balanced accuracy on a replicated study. 4. Combining geometric morphometrics with deep learning provides a reproducible, interpretable, and generalisable alternative to black-box CNN classifiers for Diptera wing analysis. By acting as a consistent single observer comparable to human annotation, the system eliminates inter-observer bias, enabling large-scale and cross-study morphometric analyses of Dipteran wings. The system is publicly available at www.ithildin.bnitm.de and transferable to other Diptera families with moderate retraining effort. Data availabilityImages used in this study are accessible under CC BY 4.0 license at https://doi.org/10.6019/S-BIAD1478. Downloadable and installable docker application can be accessed on the applications git page: https://anonymous.4open.science/r/ITHILDIN-4313/

bioinformatics↗

bioRxiv2026

NetMedGPT - A network medicine foundation model for extensive disease mechanism mining and drug repurposing

Network medicine leverages large biomedical knowledge graphs (KGs) to model disease mechanisms and identify therapeutic opportunities. However, most deep learning approaches that use KGs in biomedicine remain task-specific, limiting their ability to generalize across diverse applications within a unified framework. Here, we introduce NetMedGPT, a transformer-based foundation model trained on a large-scale biomedical KG using masked token prediction. By learning contextualized representations of biomedical nodes, NetMedGPT enables unified, zero-shot inference across different drug discovery tasks. Specifically, in five tasks, i.e., predicting the association of drugs with indications, targets, adverse drug reactions, contraindications, and off-label uses, NetMedGPT consistently outperforms all specialized baselines, achieving area under the precision-recall curve gains of between 2.2% and 26%. When evaluated on independent external datasets, NetMedGPT outperformed baseline on an expert-curated clinical indications set and also preferentially prioritized clinically relevant drug-disease pairs in ClinicalTrials.gov. NetMedGPTs generative capability further supports the construction of mechanistically plausible subnetworks offering biological insights. NetMedGPT provides a unified foundation model for network medicine that supports scalable hypothesis generation and provides potential to accelerate drug repurposing. We further provided an interactive interface (https://prototypes.cosy.bio/chatnetmedgpt/) that allows users to obtain model inferences through natural-language queries.

bioinformatics↗

bioRxiv2025

Inferring and Evaluating Network Medicine-Based Disease Modules with Nextflow

Most human diseases result from complex molecular interactions of genes and proteins. Various network-based computational methods characterize these mechanisms by expanding seed genes into disease-associated subnetworks, or disease modules. Evaluating these diverse methods is tedious due to unique installation and data preparation requirements. Moreover, the underlying algorithmic strategies differ, making it difficult to determine which of the created modules are most useful or biologically plausible. To address this challenge, we developed an all-in-one Nextflow pipeline that enables automated and reproducible analyses. It handles installation, input preparation, execution, and systematic evaluation of six widely used module detection tools, considering module topology, functional coherence, robustness, and the capacity to recover seeds. In addition, it annotates the resulting disease modules with biological context information, prioritizes potential drug candidates, and generates visualizations and a comprehensive summary report. To showcase the value of our pipeline and offer guidance to potential users, we performed a comprehensive evaluation across 50 different disease-network combinations, revealing substantial variability among the derived disease modules. We show that this variability is driven by differences in modeling approach, input network, and seed composition. While most methods are robust to minor perturbations, they struggle to recover omitted seeds, and none consistently outperforms others, underscoring the need for careful method selection. Our work enables the research community to systematically compare approaches for disease module discovery, promoting reproducible network medicine research. Integrated into the nf-core project (https://nf-co.re/diseasemodulediscovery), it is intended as an extendable, long-term resource for tracking progress in the field.

bioinformatics↗

bioRxiv2025

Aberrant splicing in human cancer has a large-scale functional impact on transcription factors

Cancer shows aberrant alternative splicing (AS), which could functionally perturb proteins required for normal cellular behavior. Transcription factors (TFs) are proteins that regulate gene transcription. Some evidence about AS-driven perturbation of TFs in cancer has been documented. A comprehensive systematic analysis of how cancer-specific AS could affect functions of TFs is missing. Such an analysis could reveal the molecular mechanisms of cancer progression due to transcription misregulations, and thus could help identify therapeutic targets. Here, we systematically analyzed potential functional perturbations of TFs due to AS across 15 cancer types. By analyzing AS patterns, we identified 2 118 perturbed AS events (i.e., events showing significant AS pattern differences between normal and paired cancer samples) that affect 718 TFs across 14 cancer types. In 189 TFs, the perturbed AS events affected known functional domains of TFs. Indirect evidence for the functional impact of cancer-specific AS on TFs were also found: First, by relating AS patterns of the perturbed AS events with the DNA-binding and regulatory activity of TFs. Second, by using cancer dependency data to explore whether the affected TFs are essential for cancer cell line proliferation. Our findings show a large-scale functional perturbation of TFs due to cancer-specific AS.

bioinformatics↗

bioRxiv2025

Development of an Integrated Single-Cell and Spatial Transcriptomics Atlas of Healthy Human Skin Focusing on the Pilosebaceous Unit

Single-cell and spatial transcriptomics have transformed our ability to chart human tissue organization at unprecedented resolution. These technologies enable the construction of high-quality reference atlases, essential for mapping healthy tissue architecture and identifying robust gene markers. We developed the healthy Human Skin Cell Atlas (HSCA), systematically integrating 34 public datasets and totaling 821,464 cells, with curated metadata and harmonized cell type nomenclature to ensure consistency. We place particular emphasis on the pilosebaceous unit, a key epithelial structure critical for both homeostasis and pathology. While prior studies captured the interfollicular epidermis and immune landscape in detail, deeper hair follicle regions remained under-characterized. By leveraging high-resolution spatial transcriptomics (Visium HD), we spatially resolved and transcriptionally defined the lower hair follicle compartments and pinpointed signalling hubs. Furthermore, the HSCA enables the detection of cell types not visible in standalone datasets, such as Merkel cells. Our results illustrate the value of the integrated single-cell atlas and spatial data in refining tissue organization and highlight the PSU as a complex and diverse epithelial niche.

bioinformatics↗

bioRxiv2025

Detection of alternative splicing: deep sequencing or deep learning?

Alternative splicing (AS) is a central mechanism of gene regulation that enables condition- and tissue-specific expression of gene isoforms. Its dysregulation plays a role in diseases such as cancer, neurological disorders, and metabolic conditions. Despite its importance, accurately identifying AS events remains challenging, especially in large-scale studies relying on publicly available RNA sequencing (RNA-seq) data. State-of-the-art AS event detection typically requires deep sequencing with over 100 million reads; however, much of the publicly accessible data is of lower sequencing depth. Recent advances, particularly deep learning models working with genomic sequences, offer new avenues for predicting AS without reliance on high sequencing depth data. Our study addresses the question: Can we utilize the vast repository of publicly available RNA-seq data for AS detection, despite often lacking the sequencing depth typically required? We show that sequence-based tools such as DeepSplice and SpliceAI show promising performance in retrieving novel and unannotated splice junctions, even when RNA-seq data are limited, but are not suitable for de novo splice junction detection. Our results demonstrate the potential of sequence-based tools for initial hypothesis development and as additional filters in standard RNA-seq pipelines, especially when sequencing depth is limited. Nonetheless, validation with higher sequencing depths remains essential for confirmation of splice events. Overall, our findings underscore the need for integrative methods combining genomic and RNA-seq data for prediction of tissue-/condition-specific AS in resource-limited settings.

bioinformatics↗

bioRxiv2025

Regulation of the transcriptome, miRNAs, and alternative splicing in a FSGS zebrafish injury model

BackgroundFocal Segmental Glomerulosclerosis (FSGS) is a severe kidney disorder with complex and not yet fully understood pathogenesis. Alternative splicing (AS) - the generation of distinct protein isoforms from the same gene - might play a critical role by the regulation of gene functions and disease development. MethodsTo investigate the role of AS in FSGS, we used a zebrafish model, which mimics key human FSGS features, including foot process effacement, matrix accumulation, podocyte detachment and parietal epithelial cell activation. We performed total RNA sequencing of isolated zebrafish glomeruli and whole larvae, followed by integrative bioinformatic analysis to identify AS events and regulatory miRNAs. ResultsOur data revealed a downregulation of essential podocyte genes (nphs1, nphs2, podxl, wt1) and an inhibition of pathways associated with nephron development and cytoskeletal organization. We also observed increased expression of the transcription factor stat3 and disease-associated miRNAs such as miR-21 and miR-193. AS analysis identified approximately [~]7,000 splicing events, primarily exon skipping ([~]80%), affecting genes such as nphs1, magi2, and ptpro. A total of 136 and 612 alternatively spliced genes were found at 5 and 6 days post-fertilization (dpf), respectively. Isoform switch analysis uncovered 70 genes affected by AS in FSGS, including epb41l5 (linked to podocyte adhesion), fgfr1a (fibroblast growth signaling), and members of the SRSF splicing factor family (e.g., srsf3a). ConclusionsThese findings emphasize the importance of transcriptional and post-transcriptional regulation, including AS, in FSGS pathogenesis. Furthermore, they support the zebrafish model as a valuable system for identifying novel mechanisms and potential therapeutic targets for kidney diseases.

cell biology↗

bioRxiv2025

Potentials and limitations in the application of Convolutional Neural Networks for mosquito species identification using wing images

1. This study addresses the pressing global health burden of mosquito-borne diseases by investigating the application of Convolutional Neural Networks (CNNs) for mosquito species identification using wing images. Conventional identification methods are hampered by the need for significant expertise and resources, while CNNs offer a promising alternative. Our research aimed to develop a reliable and applicable classification system that can be used under real-world conditions, with a focus on improving model adaptability to unencountered devices, mitigating dataset biases, and ensuring usability across different users without standardized protocols. 2. We utilized a large, diverse dataset of mosquito wing images of 21 taxa and three imagecapturing devices and an optimized preprocessing pipeline to standardize images and remove undesirable image features. 3. The developed CNN models demonstrated high performance, with an average balanced accuracy of 98.3% and a macro F1-score of 97.6%, effectively distinguishing between the 21 mosquito taxa, including morphologically similar pairs. The preprocessing pipeline improved the models robustness, reducing performance drops on unfamiliar devices effectively. However, the study also highlights the persistence of inherent dataset biases, which the preprocessing steps could only partially mitigate. The classification systems practical usability was demonstrated through a feasibility study, showing high inter-rater reliability. 4. The results underscore the potential of the proposed workflow to enhance vector surveillance, especially in resource-constrained settings, and suggest its applicability to other winged insect species. The classification system developed in this study is available for public use, providing a valuable tool for vector surveillance and research, supporting efforts to mitigate the spread of mosquito-borne diseases.

bioinformatics↗

bioRxiv2025

Systematic Evaluation of Normalization Approaches in Tandem Mass Tag and Label-Free Protein Quantification Data Using PRONE

Despite the significant progress in accuracy and reliability in mass spectrometry technology, as well as the development of strategies based on isotopic labeling or internal standards in recent decades, systematic biases originating from non-biological factors remain a significant challenge in data analysis. In addition, the wide range of available normalization methods renders the choice of a suitable normalization method challenging. We systematically evaluated 17 normalization and two batch effect correction methods, originally developed for pre-processing DNA microarray data but widely applied in proteomics, on six publicly available spike-in and three label-free and tandem mass tag datasets. Opposed to state-of-the-art normalization practice, we found that a reduction in intragroup variation is not directly related to the effectiveness of the normalization methods. Furthermore, our results demonstrated that the methods RobNorm and Normics, specifically developed for proteomics data, in line with LoessF performed consistently well across the spike-in datasets, while EigenMS exhibited a high false positive rate. Finally, based on experimental data, we show that normalization substantially impacts downstream analyses, and the impact is highly dataset-specific, emphasizing the importance of use-case-specific evaluations for novel proteomics datasets. For this, we developed the PROteomics Normalization Evaluator (PRONE), a unifying R package enabling comparative evaluation of normalization methods, including their impact on downstream analyses, while offering considerable flexibility, acknowledging the lack of universally accepted standards. PRONE is available on Bioconductor with a web application accessible at https://exbio.wzw.tum.de/prone/.

bioinformatics↗

bioRxiv2024

RNA sequencing depth guidelines for the study of alternative splicing

A key parameter in the experimental design of RNA-seq projects is the choice of sequencing depth. Considering a limited budget, one needs to find a tradeoff between the number of samples and the sensitivity of the analysis, particularly concerning lowly expressed genes. While previous studies have proposed a lower bound for the comprehensive analysis of differential gene expression, for the analysis of alternative splicing, it has only been proposed for human adipose tissue. However, alternative splicing differs across tissues and conditions. We analyzed publicly available and newly generated deep-sequenced paired-end RNA-seq samples (between 150 and >500 million reads, read length 50-150 bp) from human buffy coat cells and diverse sets of tissues, including gluteal subcutaneous fat, heart, and hypothalamus. Our results show that the sequencing depth typically used in published cohorts is not sufficient to comprehensively capture the landscape of alternative splicing. This motivates the use of deeper sequencing or long-read technologies in future studies. Toward this goal, we offer guidelines for choosing a suitable sequencing depth. GRAPHICAL ABSTRACT O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=177 SRC="FIGDIR/small/617406v2_ufig1.gif" ALT="Figure 1"> View larger version (24K): org.highwire.dtl.DTLVardef@1aecd9org.highwire.dtl.DTLVardef@1b3d0fborg.highwire.dtl.DTLVardef@5d15c4org.highwire.dtl.DTLVardef@140035b_HPS_FORMAT_FIGEXP M_FIG C_FIG

bioinformatics↗

bioRxiv2024

DREAMER: Exploring Common Mechanisms of Adverse Drug Reactions and Disease Phenotypes through Network-Based Analysis

Adverse drug reactions (ADRs) are a major concern in clinical healthcare, significantly affecting patient safety and drug development. The need for a deeper understanding of ADR mechanisms is crucial for improving drug safety profiles in drug design and drug repurposing. This study introduces DREAMER (Drug adverse REAction Mechanism ExplaineR), a novel network-based method for exploring the mechanisms underlying adverse drug reactions and disease phenotypes at a molecular level by leveraging a comprehensive knowledge graph obtained from various datasets. By considering drugs and diseases that cause similar phenotypes, and investigating their commonalities regarding their impact on specific modules of the protein-protein interaction network, DREAMER can robustly identify protein sets associated with the biological mechanisms underlying ADRs and unravel the causal relationships that contribute to the observed clinical outcomes. Applying DREAMER to 649 ADRs, we identified proteins associated with the mechanism of action for 67 ADRs across multiple organ systems, e.g., ventricular arrhythmia, metabolic acidosis, and interstitial pneumonitis. In particular, DREAMER highlights the importance of GABAergic signaling and proteins of the coagulation pathways for personality disorders and intracranial hemorrhage, respectively. We further demonstrate the application of DREAMER in drug repurposing and propose sotalol (targeting KCNH2), ranolazine (targeting SCN5A, currently under clinical trial), and diltiazem (indicated drug targeting CACNA1C and SCN3A) as candidate drugs to be repurposed for cardiac arrest. In summary, DREAMER effectively detects molecular mechanisms underlying phenotypes emphasizing the importance of network-based analyses with integrative data for enhancing drug safety and accelerating the discovery of novel therapeutic strategies.

systems biology↗

bioRxiv2024

Alternative splicing in mechanically stretched podocytes as a model of glomerular hypertension

BackgroundAlterations in pre-mRNA splicing play an important role in disease pathophysiology. However, the role of alternative splicing (AS) for podocytes in hypertensive nephropathy (HN) has not been investigated. The purpose of the Sys_CARE project was to identify AS events that play a role in the development and progression of HN. MethodsMurine podocytes were exposed to mechanical stretch, after which proteins and mRNA were analyzed by proteomics, RNA-Seq and several bioinformatic AS tools. ResultsBased on transcriptomics and proteomics analysis we could observe significant changes in gene expression and abundance of proteins under mechanical stretch compared to unstretched conditions. By RNA-Seq, we identified over 3,000 alternative spliced genes after mechanical stretch, including all types of AS events. We found 17 genes that showed an AS event in four different splicing analysis tools. From these, we focused on Myl6, a component of the myosin protein complex, and Shroom3, an actin-binding protein crucial for podocyte function. We found two Shroom3 isoforms that showed significant changes in expression upon mechanical stretch, which was verified by qRT-PCR and in situ hybridization. Furthermore, we observed an expression switch of two Myl6 isoforms after mechanical stretch. This switch is accompanied by a change in a C-terminally located amino acid sequence. ConclusionsIn summary, mechanical stretch of cultured podocytes is an excellent model to simulate hypertensive nephropathy. In depth RNA-Seq analysis disclosed alternative splicing events, such as in Shroom3 and Myl6, which may play a crucial role in the pathophysiology of hypertension-induced nephropathy.

cell biology↗

bioRxiv2024

Prognostic importance of splicing-triggered aberrations of protein complex interfaces in cancer

Aberrant alternative splicing (AS) is a prominent hallmark of cancer. AS can perturb protein-protein interactions (PPIs) by adding or removing interface regions encoded by individual exons. Identifying prognostic exon-exon interactions (EEIs) from PPI interfaces can help discover AS-affected cancer-driving PPIs that can serve as potential drug targets. Here, we assessed the prognostic significance of EEIs across 15 cancer types by integrating RNA-seq data with three-dimensional (3D) structures of protein complexes. By analyzing the resulting EEI network we identified patient-specific perturbed EEIs (i.e., EEIs present in healthy samples but absent from the paired cancer samples or vice versa) that were significantly associated with survival. We provide the first evidence that EEIs can be used as prognostic biomarkers for cancer patient survival. Our findings provide mechanistic insights into AS-affected PPI interfaces. Given the ongoing expansion of available RNA-seq data and the number of 3D structurally-resolved (or confidently predicted) protein complexes, our computational framework will help accelerate the discovery of clinically important cancer-promoting AS events.

bioinformatics↗

bioRxiv2024

Transcription factor prediction using protein 3D structures

MotivationTranscription factors (TFs) are DNA-binding proteins that regulate gene expression. Traditional methods predict a protein as a TF if the protein contains any DNA-binding domains (DBDs) of known TFs. However, this approach fails to identify a novel TF that does not contain any known DBDs. Recently proposed TF prediction methods do not rely on DBDs. Such methods use features of protein sequences to train a machine learning model, and then use the trained model to predict whether a protein is a TF or not. Because the 3-dimensional (3D) structure of a protein captures more information than its sequence, using 3D protein structures will likely allow for more accurate prediction of novel TFs. ResultsWe propose a deep learning-based TF prediction method (StrucTFactor), which is the first method to utilize 3D secondary structural information of proteins. We compare StrucTFactor with recent state-of-the-art TF prediction methods based on[~] 525 000 proteins across 12 datasets, capturing different aspects of data bias (including sequence redundancy) possibly influencing a methods performance. We find that StrucTFactor significantly (p-value < 0.001) outperforms the existing TF prediction methods, improving the performance over its closest competitor by up to 17% based on Matthews correlation coefficient. AvailabilityData and source code are available at https://github.com/lieboldj/StrucTFactor and on our website at https://apps.cosy.bio/StrucTFactor/ Contactkhalique.newaz@uni-hamburg.de Supplementary informationIncluded

bioinformatics↗

bioRxiv2024

The power and limits of predicting exon-exon interactions using protein 3D structures

Alternative splicing (AS) effects on cellular functions can be captured by studying changes in the underlying protein-protein interactions (PPIs). Because AS results in the gain or loss of exons, existing methods for predicting AS-related PPI changes utilize known PPI interfacing exon-exon interactions (EEIs), which only cover [~]5% of known human PPIs. Hence, there is a need to extend the existing limited EEI knowledge to advance the functional understanding of AS. In this study, we explore whether existing computational PPI interface prediction (PPIIP) methods, originally designed to predict residue-residue interactions (RRIs), can be used to predict EEIs. We evaluate three recent state-of-the-art PPIIP methods for the RRI- as well as EEI-prediction tasks using known protein complex structures, covering [~]230,000 RRIs and [~]27,000 EEIs. Our results provide the first evidence that existing PPIIP methods can be extended for the EEI prediction task, showing F-score, precision, and recall performances of up to [~]38%, [~]63%, and [~]28%, respectively, with a false discovery rate of less than 5%. Our study provides insights into the power and limits of existing PPIIP methods to predict EEIs, thus guiding future developments of computational methods for the EEI prediction task. We provide streamlined computational pipelines integrating each of the three considered PPIIP methods for the EEI prediction task to be utilized by the scientific community.

bioinformatics↗