Schedule

The Bioconductor Training Committee has had a successful past year and many plans for the next. One of our main accomplishments, Carpentries-style workshop materials, were used for in-person workshops covering Intro to R and Bulk RNA-Seq in Kenya, Ethiopia and Benin, while the Bulk RNA-Seq and Single Cell RNA-Seq workshops were offered before EuroBioC2025 in Spain. This year a new workshop, Microbiome Analysis, is being developed and will be held before EuroBioC2026 in Finland. Additionally, the Training Committee has partnered with Physalia for online workshops taught by experienced members of the Bioconductor community, with two free seats available for participants from under-represented countries. As a follow-up to our recent publication on “Learning and teaching biological data science in the Bioconductor community” (Drnevich, Tan et al. 2025), we are exploring the development of an ELIXIR Training eSupport System (TeSS) Space for a centralised, Bioconductor-focused view of training materials, videos, event listings, and specific “learning paths” highlighting which workshops, package vignettes and/or other training materials would be useful for a particular analysis goal. Committee member Kevin Rue-Albrecht has been awarded a Software Sustainability Institute Fellowship for “Improving Software Discoverability in the Life Sciences - Building Bridges between Bioconductor and the EDAM ontology” and also a BioFAIR Pathfinder Project “Connecting Bioconductor, Galaxy, and nf-core: FAIR, AI-ready workflows and training for single-cell analysis.” Both of these initiatives will strengthen connections between Bioconductor and two other major bioinformatics platforms to improve the visibility, accessibility, and reuse of complementary software and training materials across life science communities. We also consider how the increasing use of AI tools by learners may affect training, including their use for code generation and pair programming, and how to ensure that this does not replace core learning activities or undermine skills in statistical reasoning, reproducibility, and critical evaluation of results. Finally, the Training Committee developed governance documents to formalize co-chair elections and terms, and define other members’ roles. We encourage any interested community member to join the Training Committee’s monthly meetings and help out however they can.

The Bioconductor project has recently benefitted from direct funding for community and developer engagement roles, as well as indirect support for community and developer events homed in grants to Bioconductor contributors and core team members. The abstract goal of this direct and indirect support is to ensure that the Bioconductor project continues to provide robust, replicable, and relevant software tools for the scientific community. The challenges facing the Bioconductor project are not unique. Project contributors are already invested in providing novel work, but soliciting volunteer work on project infrastructure, critical maintenance, and forward-compatibility of project systems from contributors can have mixed success. Developer and community engagement roles are intended to address these challenges directly by running official events, formalizing community contributions to infrastructure, and emphasizing and advertising community contributions to the project. This talk will provide an overview of recently completed and upcoming community events, an introduction to the general mandates of the community development and developer engagement roles, and the intended structures that these roles intend to use to coordinate, and facilitate community contributions to the Bioconductor project.

Within Bioconductor there are 72 packages that list flow cytometry in their BiocViews, collectively permitting analysis of conventional, spectral and mass cytometry data. However, the majority of the packages are underutilized, with only 15 surpassing 4000 yearly downloads. A contributing factor to this is that traditionaly, flow cytometry analysis is carried out using commercial software (ex. FlowJo and FCSExpress), with a graphical-user interface being used to draw gates around a cell population of interest on 2-D plots. Cells falling within this gate are filtered for, additional gates are subsequently drawn, resulting in a gating hierarchy enabling the isolation of a cell population of interest for statistical analysis. With the emergence of spectral flow cytometry in the last decade, which is capable of profiling 20-50 markers on millions of cells within a few minutes, the resulting datasets are growing increasingly complex. Semi-supervised and unsupervised analytical methods (often implemented as Bioconductor packages) will need to be used in the analysis. For most flow cytometrist, with no-to-limited coding skills, this presents a barrier to entry. While learning resources exist for those with intermediate coding skills (such as individual package vignetters, workflows, and the odd workshop recording), these are not aimed at beginners. Consequently, many self-study attempts end in frustration. To address this urgent community need (and reduce the barrier to entry), starting in February 2026 we have been offering a free weekly “Cytometry in R” course, offered both in-person and online, aimed at flow cytometrists with no coding experience. The community response was outstanding, with over 1993 worldwide participants filling out the interest form, 529 participants creating new GitHub accounts and forking the course repository, and between 300-500 distinct viewers per weekly topic. This course covers one topic per week, with multiple livestream offerings to accomodate different timezones. Recordings are available via our YouTube channel (https://www.youtube.com/@CytometryInR). The course is run out of a GitHub repository (https://github.com/UMGCCCFCSR/CytometryInR), with all course materials, code and datasets being offered under CC-BY-SA and AGPL3-0 licenses. This permits hosting the course website (built using Quarto) as a GitHub page, and utilization of the Discussion page as a community forum to field beginner questions ranging from installation errors, missed library calls, etc. In addition to teaching R fundamentals to beginners within a cytometry-focused context, we also reinforce how to use Git for version control, Quarto for reproducible documentation, and other good coding practices. The course is scheduled to continue for 30 weeks, gradually covering intermediate and advanced content (https://umgcccfcsr.github.io/CytometryInR/Schedule). All recordings and course materials will remain available after the course concludes, providing a framework that future self-learners can utilize to make their learning journeys smoother than the ones we experienced when first getting started. We highlight what has worked, things we wish we had done differently, and unappreciated elements encountered while reducing barriers to entry to a community that would “rather die than ever touch a command line”. We hope the resources made available through the course enable wider utilization of Bioconductor flow cytometry packages, and inspire next generation of package maintainers and developers.

Recent reductions in sequencing cost have made complex experimental designs much more accessible, making it routine for transcriptomic or proteomic assays to yield dozens of interesting comparisons. Biological context might further warrant pattern analysis to find co-regulated genes, functional enrichment analysis to detect pathways affected by perturbation or orthogonal assays conducted in parallel, e.g. proteomics or mass spectrometry. In addition, tool parameters often need to be fine-tuned over multiple rounds to maximize statistical power and facilitate biological interpretation. In the end, the massive volume of scripts, output files and analysis objects can be overwhelming to manage not only for biologists but even for computational experts, making the task of extracting biological insight even more challenging. To address this we developed Carnation, an R/Shiny package that integrates various facets of differential expression analysis into a unified, interactive environment designed for both computational and experimental researchers. Carnation offers a clean intuitive graphical interface for exploring differential expression results, clustering patterns, and functional enrichment summaries, supporting multiple modalities and implementing several interactive features, such as highly customizable visualizations, and a global scratchpad for keeping track of genes of interest. Carnation features a modular extensible design allowing easy integration with new tools or visualizations and can be deployed locally or as a sharing platform on public infrastructure. In summary, Carnation democratizes multi-omic genomics analyses, transforming static analysis into a collaborative, interactive experience, enabling research teams to spend less time navigating data and more time generating biological insight.

Reproducibility and scalability are central challenges for many bioinformatics analyses. These analyses often involve specific versions of software packages, system libraries, and analysis scripts that can break or produce different results when anything changes. Scaling from a few samples to hundreds typically requires ad hoc solutions that are fragile and hard to maintain, making it far harder to reproduce an analysis than it should be. Workflow languages address this by defining analyses as portable, version-controlled scripts paired with containerized environments. Our team at Fred Hutchinson Cancer Center has adopted the Workflow Description Language (WDL) for its readability and approachable learning curve. However, the upfront investment of writing workflows from scratch remains a significant barrier for many researchers. To lower this barrier, we developed the WILDS WDL Library, a rigorously tested, well-documented collection of modular WDL workflow scripts. The scripts are organized into two tiers: modules, which wrap popular commands for over 40 bioinformatics tools (including several Bioconductor packages) into reusable tasks, and pipelines, which are pre-written end-to-end workflows that use the module tasks. Each resource is containerized, tested across multiple workflow executors, and designed to work out of the box. At Fred Hutch, these scripts have reduced the barrier to entry for researchers and enabled them to harness parallelization on the cluster, in some cases cutting computational wait times from weeks to hours. We are continuing to expand the library, which could be used to make Bioconductor-powered analyses more reproducible and scalable, regardless of a researcher's computing environment.

As AI-driven computational biology advances, the primary bottleneck for organizations often shifts from model development to infrastructure scaling. Complex workloads, such as protein structure prediction, require highly scalable, automated, and reproducible cloud environments. This session explores how to architect and orchestrate these compute-intensive pipelines using cloud-native infrastructure with SimpleFold, an efficient, transformer-based prediction model utilizing Apple’s MLX framework, as our real-world use case. We will walk through the end-to-end lifecycle of deploying AI workflows at scale in the cloud. We will explore how to encapsulate complex model dependencies using containerized Docker pipelines and how to automate the underlying infrastructure deployment using Terraform. Crucially, we will highlight how AWS HealthOmics serves as the workflow orchestration engine, providing a managed, scalable environment designed to run reproducible bioinformatics pipelines without the operational overhead of managing underlying compute clusters.

Presentation slides

This talk will introduce two relatively new and less known binary distribution initiatives from the Bioconductor Core Team, namely WebR-based Bioconductor builds available at webr.bioconductor.org for entirely client-side use of Bioconductor packages directly in the browser, and Bioc2u packages for easy access to pre-compiled binaries and system dependencies on Debian/Ubuntu systems. In this talk, the audience will be familiarized with what these binaries are, how to use them, and how they make Bioconductor software faster and easier to install and deploy on these platforms. I will briefly explain the design and infrastructure behind these binaries, then walk through practical examples: using Bioconductor interactively via WebAssembly in the browser, installing Bioconductor packages with apt on Ubuntu, and explore potential usecases integrating these workflows into teaching, containers, and reproducible analysis pipelines.

This workshop gives a quick overview of the Bioconductor project, particularly the shared infrastructure that facilitates inter-operability between Bioconductor packages. The hands-on part will explore one of the most commonly-used S4 objects, a SummarizedExperiment. We will review it’s structure, how to manipulate it and make figures, and how to construct one from your own data. Finally, we will briefly explore the Bioconductor package repository and learn how to install Bioconductor on your own computer.

R Shiny modules comprise paired UI and server functions that communicate through shared namespaces (NS()), ensuring proper isolation, reusability, code organization, and testability. InteractiveComplexHeatmap is a Bioconductor package that enables interactive visualization of complex heatmaps within Shiny applications. This work introduces a Shiny module implementation for the InteractiveComplexHeatmap package and demonstrates its integration with Teal, a mature meta-UI framework specifically designed for clinical and genomic data exploration. By assembling these advanced tools with minimal configuration, developers can rapidly create sophisticated, user-friendly dashboards for common genomic data visualizations. The Teal framework provides an envelope layer that manages universal data filtering, session management, and code reproducibility features automatically, significantly reducing the customization burden on application maintainers while delivering enterprise-grade functionality. Key contributions: Modularization of InteractiveComplexHeatmap for enhanced reusability; demonstration of streamlined integration with Teal framework; simplified pathway to production-ready genomic data exploration tools. Easy assembly with minimal configuration for state-of-the-art developer and user tested app features.

Accurate species-level classification of microbial communities remains a major challenge in microbiome analysis, particularly when using traditional 16S rRNA amplicon sequencing pipelines as many common methods often fail to resolve taxonomy beyond the genus level. To address this, we present MetaScope, a modular, R-based software package that reimplements and extends the PathoScope 2.0 framework for high-resolution microbial classification. MetaScope introduces two key innovations: (1) the integration of user-defined or empirical prior weights into PathoScope’s Bayesian read reassignment algorithm to improve abundance estimation, and (2) MetaBlast, a secondary BLAST-based validation module for refining species-level assignments. Benchmarking against mock and clinical datasets demonstrates that MetaScope significantly outperforms common existing methods such as QIIME2 and DADA2, achieving up to 93.6% species-level classification accuracy which translates too enhanced resolution for downstream ecological analyses. These results highlight MetaScope as a powerful tool for advancing microbial community profiling in both research and clinical settings.

Batch effect correction is a common and often necessary step in -omics and imaging data analysis to reduce bias due to technical and experimental factors when combining multiple batches of data. The severity of the batch effects dictates the correction strategy; therefore, a careful assessment of each dataset's batch effects is necessary. BatchQC is an R package that provides reproducible tools and visualizations for quantitatively and qualitatively addressing batch effects across a broad range of data types. BatchQC integrates with standardized Bioconductor data structures and features an object-oriented design, enabling the application of workflows that can freely evaluate and process data within and outside the package tools. Common batch evaluation methods, such as clustering, kBET, and differential expression, along with novel quantitative metrics, help determine the benefits of batch correction for each dataset and enable direct comparisons between methods. Additionally, BatchQC employs methods that assess the distributional fit of several commonly used distributions for -omics data. Importantly, these goodness-of-fit methods are not only relevant to batch effect evaluation but also apply to other analytical scenarios, like differential expression. Here, we present BatchQC as the first comprehensive batch-correction R package, with independent tools, reproducible workflows, visualization, and novel statistics.

High-plex proteomics platforms are generating increasingly rich datasets, but researchers often lack dedicated, open-source tools for the full analytical workflow from raw instrument output to publication-ready results. NULISAseq (Feng, Beer, Hao et al., Nature Communications, 2023) is a multiplexed proteomic assay that uses dual-selection proximity ligation with next-generation sequencing readout to quantify hundreds of proteins simultaneously with attomolar sensitivity and a dynamic range spanning 12 orders of magnitude. The platform supports both relative quantification (RQ) and absolute quantification (AQ) assay modes. We present NULISAseqR, an open-source R package that provides a comprehensive analysis framework for NULISAseq data. The package handles the complete workflow: XML data import and multi-plate merging, plate, sample, and target-level quality control with automated HTML reporting, intra-plate and inter-plate normalization, limit-of-detection and detectability assessment, differential abundance testing via linear and mixed-effects models, single-protein outcome modeling (linear, logistic, and mixed-effects regression), and visualization including heatmaps, PCA, and volcano plots. NULISAseqR integrates with Bioconductor infrastructure including ComplexHeatmap and PCAtools and is accompanied by a comprehensive bookdown user guide with case study workflows. NULISAseqR is available under the GPL-3 license at https://github.com/Alamar-Biosciences/NULISAseqR with full documentation at https://nulisaseqr.alamarbio.com.

Adaptive immune receptor repertoire (AIRR) sequencing has become a powerful approach for profiling B and T cell diversity in health and disease. A key step in AIRR-seq analysis is the assignment of V, D, and J germline genes to immunoglobulin and T cell receptor sequences, a task for which IgBLAST is one of the most widely adopted tools. However, integrating IgBLAST into R-based bioinformatics workflows has historically required manual installation, command-line configuration, and custom parsing of outputs — creating friction for researchers working primarily in R/Bioconductor. igblastr is a new Bioconductor package that addresses these challenges by providing a seamless, user-friendly interface to a local IgBLAST installation directly from R. The package automates IgBLAST installation from NCBI's official binaries, manages germline database setup for multiple organisms including IMGT-sourced references, and exposes a clean R API for running igblastn analyses with minimal configuration. In this talk, we will present the package design and walk through a typical workflow, from installation to germline database selection and sequence annotation.

Trisomy 21 (T21), resulting in Down Syndrome, is the most common chromosomal abnormality worldwide and is associated with multiple developmental defects, including lung anomalies. Prenatal T21 lungs exhibit abnormal dilatation of the distal airways and acini, consistent with impaired epithelial structure; however, the transcriptional and regulatory basis of epithelial dysfunction during lung development remains poorly understood. We analyzed integrated single-nucleus (sn)RNA-seq and chromatin accessibility data from 23 prenatal lung samples (14 T21, 9 D21; 13-20 weeks gestational age). Following SoupX decontamination, stringent quality control, and consensus doublet removal, RNA data were normalized using SCTransform, integrated via RPCA, and clustered in Seurat, while chromatin accessibility data were processed using Signac. Epithelial nuclei were subsetted, reprocessed, and subclustered (resolution 0.5), and annotated using curated marker genes. Differential expression was performed using NEBULA, modeling raw counts with karyotype, age, and sex as fixed effects and subject as a random effect. Chromatin accessibility analyses identified differentially accessible regions, transcription factor motif enrichment, and peak-to-gene regulatory associations. Pathway-level changes were assessed using over-representation analysis (GO) and gene set enrichment analysis (GSEA). All analyses were performed in R within a reproducible multi-omic analytical framework. Integrated transcriptomic and chromatin accessibility analyses identified disruption of epithelial adhesion programs in prenatal T21 lungs. T21 samples demonstrated impaired epithelial junctional integrity, supported by decreased expression of tight junction components such as OCLN and ultrastructural evidence of junctional disorganization. Across developmental comparisons (<15 weeks, >16 weeks no anomalies, and >16 weeks anomalies), adhesion-associated gene programs showed differential expression patterns, with corresponding chromatin accessibility changes at associated regulatory regions. Differential accessibility and motif enrichment analyses indicated regulatory changes linked to adhesion-associated pathways rather than global chromatin remodeling. RT-qPCR validation confirmed differential expression of key adhesion and junction-related genes, including MYO9B, CXADR, CADM1, PPFLA, and ITGB1, across T21 and D21 comparisons. Adhesion dysregulation emerges as a central feature of prenatal T21 lung epithelium, supported by coordinated transcriptional and regulatory alterations in epithelial adhesion programs. Together, these results demonstrate how integrated multi-omic analysis enables resolution of regulatory mechanisms underlying epithelial dysfunction in developmental disease.

Computational methods are essential tools for modern microbiome research. Yet, challenges such as lack of standardization, reproducibility, and transparency often limit reliable analysis and interpretation. Bioconductor addresses these challenges through a global, community-driven network that provides a robust, open-source ecosystem of high-quality tools. The presentation highlights how collaborative development supports microbiome research. In this domain, a growing ecosystem of packages builds on shared Bioconductor data structures such as TreeSummarizedExperiment, which extends the widely adopted SummarizedExperiment class. By relying on interoperable classes and consistent design principles, tools developed by different groups can work seamlessly together. For users, this reduces time spent on data wrangling, broadens access to well-established statistical methods, and facilitates the application and benchmarking of approaches developed across different fields. The collaborative nature of the ecosystem ensures that methods and workflows continue to advance through contributions and shared expertise from a diverse community of developers and users. This common foundation enables robust and evidence-based analysis workflows while supporting methodological innovation. It lowers barriers for collaboration between package authors, encourages method reuse, and accelerates cross-disciplinary exchange. As microbiome research progresses toward multi-omics integration, longitudinal studies, and other increasingly complex study designs, interoperable infrastructure supported by an engaged community becomes increasingly critical for scalable and reproducible analyses. Orchestrating Microbiome Analysis (OMA) online book

Multimodal physiological signals---heart rate variability, peripheral perfusion, and vascular phase coherence---contain discriminative information for neurovascular instability~(NVI) prior to structural pathology, and are accessible from wearable-compatible PPG sensors, as validated across three independent public datasets. \textit{Background:} Neurovascular functional instability~(NVI)---the pre-structural dysregulation of cerebrovascular autoregulation preceding overt stroke---represents a critical yet underserved target for community-based screening. \textit{Methods:} We propose and computationally validate the design of Melaguard, a wearable biosensing framework intended to integrate: (i)~a hydration-activated PHBV:eumelanin composite patch~(60:40~w/w) enabling skin-tone-equitable optical and bioimpedance sensing; (ii)~a Transformer-lite classifier~(1.2M parameters) fusing four NVI modalities---SpO$_2$, heart rate variability~(HRV), microvascular perfusion index~(PI), and bilateral phase coherence---into a composite NVI Score~(0--100); and (iii)~AeroKernel, a POSIX-compliant microkernel providing deterministic, privacy-by-design edge execution. Validation proceeded in three stages: synthetic simulation ($n{=}10{,}000$), clinical cohort validation on PhysioNet CVES ($n{=}172$; 84~ischemic stroke, 88~controls), and signal processing pipeline validation on PhysioNet BIDMC ($n{=}53$ ICU recordings). \textit{Results:} On synthetic data, Transformer-lite achieved AUC$\,{=}\,0.88$ [95\%\,CI: 0.83--0.92], sensitivity$\,{=}\,0.92$, specificity$\,{=}\,0.88$. On the CVES cohort, HRV-SDNN significantly discriminated stroke from control ($96.1{\pm}28.6$\,ms vs.\ $110.5{\pm}29.5$\,ms, $p{=}0.011$; Cohen's $d{=}0.49$), cross-validated by waveform-extracted HRV ($p{=}0.038$). The Transformer-lite model achieved AUC$\,{=}\,0.755$ [0.630--0.778] on real clinical data, outperforming LSTM~(0.643), Random Forest~(0.665), and SVM~(0.472). BIDMC pipeline validation confirmed PPG-derived pulse rate ($r{=}0.748$, $p{<}0.001$) and HRV surrogate (PRV $r{=}0.690$, $p{<}0.001$; Bland--Altman bias\,${=}{+}14.5$\,ms) against clinical ground truth. \textit{Conclusions:} Melaguard demonstrates computational feasibility and real-data biological validity for multimodal NVI screening. Material-level melanin biosensing is validated computationally; a clinical pilot study is planned. These results support a broader shift from episodic, imaging-based diagnosis toward continuous physiological monitoring for early stroke risk stratification.

Single-cell RNA sequencing projects often span months or years, producing dozens of .rds files whose origins, conditions, and processing history become impossible to recall over time. Opening an old Seurat or SingleCellExperiment object a year later raises immediate questions: Where did this come from? What animal? What condition? What processing steps were applied? seuratPassport solves this by stamping the object itself with a permanent, self-contained passport; so the object always remembers, even when you don't. seuratPassport is a Bioconductor R package that embeds a metadata passport directly inside Seurat, SingleCellExperiment, and SummarizedExperiment objects. The passport travels with the object forever inside the .rds file and requires no external files, no spreadsheets and no lost notes. Users can record animal info, experimental conditions, project details, lineage (parent/child relationships between subsets), RDS registry numbers, and unlimited custom fields. The package provides three core functions: seuratPassport() opens an interactive Shiny gadget to fill or update the passport; read_passport() prints the full passport and processing history to console; and log_step() appends processing steps with timestamps and cell counts. Additional features and improvements to existing functionality are currently under development. seuratPassport is under Bioconductor review for release in Bioc 3.21

Motivation: Oncofetal (OnF) gene reactivation drives tumor plasticity and therapy resistance in colorectal cancer (CRC), yet its spatial logic across primary and metastatic sites remains unknown. Most scRNA-seq workflows lose tissue context, while spatial methods often lack single-cell resolution. We present a fully reproducible R/Bioconductor pipeline that integrates multi-region scRNA-seq and Visium spatial transcriptomics (ST) to map OnF states within the tumor microenvironment (TME) and test their regulatory dependency on AP-1 transcription factors (FOS/JUND).

Methods: We analyzed a novel public CRC dataset (GSE261004: 16 PDOs & 8 matched biopsies; GSE280318: Visium HD from 4 primary tumors + 4 liver metastases). scRNA-seq was processed with Seurat (QC, SCTransform, integration), annotated via SingleR (celldex), and OnF signature (OnFS) scored with AddModuleScore. Trajectory inference used Monocle3 and tradeSeq. ST data were loaded into SpatialExperiment, normalized with nnSVG for spatially variable genes, and deconvolved using SpatialDWLS (R-native). Spatial niches were identified with BayesSpace. AP-1 regulon activity was inferred via dorothea + decoupleR on both single-cell and spatial data.

Results: scRNA-seq revealed that OnFS-high cells are enriched in metastases and co-express fetal genes (IGF2, DLK1, HMGA2) alongside AP-1 targets. BayesSpace clustering of ST data localized OnFS-high spots to hypoxic invasion fronts and immune-excluded perivascular niches in metastases, but not in primaries. decoupleR showed that FOS/JUND regulon activity is suppressed in OnFS-high spots, suggesting AP-1 loss enables fetal reprogramming. Ligand-receptor inference (NicheNet) identified CXCL12+ fibroblasts and SPP1+ macrophages as potential niche-supporting cells. Spatial trajectory analysis with tradeSeq mapped a continuous gradient from OnFS-low to OnFS-high along the invasive edge.

Conclusion: Oncofetal reprogramming in CRC is spatially organized, metastasis-enriched, and linked to AP-1 silencing. Our R-first, containerized workflow (Docker + BiocManager) is openly available and enables reproducible spatial mapping of any gene signature. This advances precision oncology by identifying targetable spatial hubs (e.g., CXCL12+ niches) and provides a template for integrating scRNA-seq + ST in R/Bioconductor.

Availability: Code & tutorials: github.com/Over89Dose/CRC_oncofetal_spatial

Keywords: Spatial transcriptomics, single-cell RNA-seq, colorectal cancer, oncofetal reprogramming, Bioconductor, BayesSpace, decoupleR, tradeSeq

Protein post-translational modifications (PTMs) in the cytoplasm and nucleus serve as direct molecular readouts of cell signaling pathways that govern cell differentiation, cancer and other diseases. However, interpreting raw PTM data into pathways requires complex computational steps that may not be accessible to all researchers. We present the R package PTMsToPathways (P2P) which implements previously-validated methods to integrate PTM proteomic data with protein-protein interactions (PPIs) and cell signaling pathways to increase access to this analysis pipeline. The strategy clusters PTMs to reveal patterns in their expression under different experimental conditions. These clustering patterns are then used to filter protein-protein interactions from external PPI databases and pathway-pathway interactions from BioPlanet’s curated sets of proteins. The resulting networks reveal data-supported interactions among both individual proteins and groups of proteins representing well-defined cell signaling pathways. These networks can be interrogated by investigators to facilitate tailored, hypothesis-driven investigation and exploratory data analysis. P2P also enables creation of consistent, visually interpretable graphs in Cytoscape, an open source network visualization application. To promote usability, the package will be submitted to Bioconductor.

In many Bioconductor pipelines, BAM access is still a major bottleneck. The core bottleneck is not only BAM decompression itself, but also the fact that existing R-facing workflows often rely on effectively single-threaded per-file access patterns. BamScale addresses this by exposing OpenMP threading from the ompBAM engine directly at the R interface, while still returning Bioconductor-friendly objects.

The practical design goal is not to replace familiar Bioconductor workflows with a separate ecosystem. Instead, BamScale aims to preserve the way users already work with:

Rsamtools::scanBam()-style field extraction, GenomicAlignments::readGAlignments()-style alignment-object workflows, BiocParallel file-level execution when multiple BAMs are processed together. This means BamScale is intended to fit where existing tooling already fits, but with an additional within-file threading axis that can remove the current parsing bottleneck for alignment-centric workloads.

Large-scale genomic, proteomic, and transcriptomic datasets can be explored through the use of functional enrichment analysis. This approach identifies functional terms that are significantly over-represented in a query by integrating information from biological databases such as Reactome and Gene Ontology. To support this analysis, multiple tools have been developed, such as DAVID, clusterProfiler, g:Profiler, and Enrichr.

The Bioconductor enrichViewNet package enables the visualization of functional enrichment results as biological network graphs. Network-based visualization provides a distinct advantage over traditional representations (bar charts and dot plots) by highlighting relationships and potentially uncovering underlying biologically relevant patterns. Two types of network graphs can be created: customizable gene-term networks and enrichment maps. The package works with results from the functional enrichment gprofiler2 package, an R client for the g:Profiler web server – chosen for its seamless integration, widespread adoption, and continuous updates to its knowledge database.

The first type of network graphs generated by the enrichViewNet package is a gene-term network. In it, genes and functional terms are represented as nodes, with edges connecting genes to their functional term. These networks can be loaded directly into Cytoscape software with the Bioconductor RCy3 package.

In contrast, enrichment maps show enriched terms as nodes, connected by edges when these share a minimum ratio of common elements. So, the intricate relationships between functional terms are highlighted. In addition, enrichment results from multiple functional enrichment analyses can be visualized on the same graph.

In the most recent 2026 release of the enrichViewNet package, the enrichment maps are now generated as igraph objects, a data structure that encapsulates the mathematical properties of a graph, as created by the CRAN igraph package. This format increases the user's ability to personalize the final visual output.

To illustrate the enrichViewNet package capabilities, a vignette showcasing differences in transcriptomic expression in 3D cultures of pancreatic stellate cells, most specifically related to axon-related pathways, will be included. This highlights the role of the nervous system in pancreatic ductal adenocarcinoma (PDAC) and its interaction with the tumor microenvironment. In addition, we analyzed a transcriptomic dataset from four 2D PDAC cell lines with varying levels of resistance to RAS inhibitors.

In summary, the enrichViewNet package simplifies functional enrichment analysis by converting results into intuitive biological networks. It integrates seamlessly into an R workflow and empowers researchers to uncover and interpret key biological insights.

The enrichViewNet package is available on Bioconductor: https://bioconductor.org/packages/enrichViewNet

To drive a major increase in the availability of human molecular data for genetic ancestry-informed research, we have recently developed an R package RAIDS (Robust Ancestry Inference using Data Synthesis). The initial Bioconductor release of RAIDS in the fall of 2023 provided functionality for inferring global genetic ancestry at a continental level of resolution from sequence data for which this purpose is challenging. These included RNA-seq, small panels targeting a few hundred genes, and whole-genome or whole-exome sequencing of DNA with severe somatic alterations caused by cancer. A key distinguishing feature of RAIDS is its ability, specifically for given sequence data from a given human nucleic acid donor, to tune the inference parameters for optimal performance and to assess inference accuracy at the optimum. Underlying this ability is a novel technique for synthesizing read data with any given coverage and read quality, with known, "ground-truth", genotypes throughout the donor's genome.

As a result of subsequent development, RAIDS has undergone a major upgrade (soon to be available in Bioconductor) with expanded options for inferring genetic ancestry from read data originating from ATAC-seq, single-cell RNA-seq, and ChIP-seq. Furthermore, RAIDS now enables inference of ancestral admixtures, for a more in-depth characterization of the donor's ancestral background. In parallel, we are updating the reference data to increase the specificity of the ancestry by adding the Human Genome Diversity Project to the 1000 Genomes Project. Finally, we are developing a database paired with a visual interface to facilitate data management and enhance accessibility.

The RAIDS package is available on Bioconductor: https://bioconductor.org/packages/RAIDS/

KRAS G12D is one of the most prevalent oncogenic drivers in pancreatic and colorectal cancers and remains a challenging therapeutic target because of its high affinity for GTP and the absence of a reactive residue for covalent inhibition. Identifying non-covalent inhibitors capable of targeting multiple conformational states provides a strategy to overcome these limitations. Methods: An open-source computational workflow integrating virtual screening, molecular docking, AI-guided ligand optimization using DeepFrag, and ADMET prediction was employed to identify and optimize potential dual-state KRAS G12D candidate binders. FDA-approved compounds from the ZINC World-approved library were screened against inactive (7RPZ) and active (7T47) KRAS G12D crystal structures. Promising candidates were structurally optimized and subsequently evaluated for binding affinity, binding interactions, and pharmacokinetic properties. Results: Virtual screening identified split1062 as a promising dual-state scaffold suitable for AI-guided optimization. Three derivatives were generated; among them, opt3 retained favorable docking scores against both conformational states while improving its drug-like properties. Interaction analysis revealed conserved hydrogen bonding with the mutant-Asp12 residue in both conformational states. ADMET profiling indicated opt3 satisfied Lipinski's Rule of Five, exhibited high gastrointestinal absorption, lacked PAINS and Brenk alerts, and eliminated the predicted blood-brain barrier permeability observed for the parent compound while maintaining favorable drug-like properties. Conclusion: This study suggests integrating AI-guided ligand optimization with virtual screening can optimize the pharmacokinetic properties while retaining predicted binding and pharmacokinetic properties of repurposed predicted KRAS G12D. The optimized compound opt3 displayed favorable dual-state binding characteristics together with an improved predicted drug-likeness profile.

Visium HD is a relatively recent spatially resolved transcriptomics platform released by 10x Genomics in 2024. Before its introduction, researchers typically had to choose between genome-wide measurement (with platforms like Visium standard) or high enough resolution to measure individual cells (“subcellular resolution”) with platforms like Xenium, with neither platform having both features. In contrast, Visium HD measures the spatial locations of thousands of protein-coding genes at subcellular resolution, introducing new analysis possibilities that are in active development within the research community. Quantifying gene expression within subcellular 2um-wide bins, Visium HD enables not only the quantification of nuclear transcripts, but those in the cell body or even the extracellular space, an area previously shown in the brain to be relevant in certain psychiatric disorders. This talk aims to enumerate several possibilities for leveraging the extracellular space in Visium HD data to answer biological questions that are not typically tractable in other assays. Our team has access to a five-sample human brain dataset, where we’ve explored analysis options in R/Bioconductor and Python. After first defining the extracellular space, we talk about clustering approaches, quantifying risk for phenotypes like psychiatric disorders, and even a notion of cell-cell communication in the extracellular space. GitHub

Introduction Tissue function emerges from coordinated signaling between cells, yet capturing these interactions in their native spatial context has long been technically out of reach. Imaging-based spatial transcriptomics (iST) addresses this by delivering gene expression profiles alongside precise single-cell coordinates, enabling physically grounded cell-cell interaction (CCI) inference at scale. Despite the advantage, appropriate normalization to mitigate technical variability remains a key analytical challenge (Guo et al., 2026). Library size (LS) in ST does not solely capture technical variability but also reflects underlying biology (Bhuva et al., 2024). Particularly with iST, the measurements are shaped by substantial technical variability arising from the use of targeted gene panels, transcript capture efficiency, and cell segmentation quality that, if uncorrected, can distort the expression patterns underpinning CCI inference. Most current workflows apply normalization methods from single-cell RNA sequencing, which do not account for spatially structured variation in gene expression, a feature unique to iST (Atta et al., 2024; Crowell et al., 2025). To address this, spatially aware normalization methods such as SpaNorm have been proposed, which decomposes spatially smooth variation into biological signal and technical components driven by LS (Salim & Bhuva, 2025). These distinct normalization strategies differ fundamentally in their underlying assumptions, yet the impact of this choice on CCI inference remains poorly understood (Atta et al., 2024; Bhuva et al., 2024; Salim & Bhuva, 2025). Understanding the impact of normalization on CCI analysis is important because normalization can suppress or inflate L-Rs co-expression patterns that CCI methods rely on, potentially leading to spurious or missed interactions. To address this, we systematically evaluated the impact of 11 normalization methods on CCI inference using a benchmark of well-annotated iST datasets spanning multiple platforms and tissue contexts. Methods We curated thirteen iST datasets that cover seven iST technologies with various gene panel sizes, seven tissue types, and two species. The platforms included 10× Genomics Xenium In Situ, Nanostring CosMx Spatial Molecular Imager, Vizgen MERSCOPE, Xenium-5K, CosMx-6K, CosMx-Whole Transcriptome (WTX), and MERSCOPE-1K. The datasets cover lung, breast, brain, colon, pancreas, liver, and ovaries tissues across human and mouse. Eleven normalization options were evaluated, including no normalization, log transformation, DESeq2, Seurat (LogNormalization, relative counts (RC), and centered log ratio transformation (CLR)), scran, Giotto, sctransform (SCT), RUV-III-NB, and SpaNorm. CellChat was used to detect spatially proximal CCIs (Jin et al., 2025). The number of L-Rs and their interaction strength were examined to characterize the overall CCI landscape. CCIs consistency across normalizations in each sample was evaluated using L–Rs shared by at least half of the methods as an alternative positive set, and F1 scores were computed, since ground-truth L–Rs are unavailable. Results A total of 151 comparisons were conducted. Overall, fewer and weaker L-Rs were predicted from the normalized CosMx-WTX and MERSCOPE-1K data compared to that of the other platforms. Specifically, the median number of L-Rs in the CosMx-WTX human pancreas, CosMx-WTX human colon, and MERSCOPE-1K human brain samples were 2, 3, and 36 respectively, whereas for other datasets it ranged from 162 to 4,156. Similarly, the median interaction strength was 0.003, 0.0002, and 0.273 in the three samples, compared to 1.362 to 63.682 in the remaining datasets. Within each sample, different normalization options had little effect on the number of predicted L-Rs, however they did influence CCI strength, with LogNormalization, Giotto, and SpaNorm ranking among the top three in 10 out of 13 datasets, producing stronger CCIs than the other normalizations within each dataset. Moreover, L-Rs predicted from RC-, scran-, and CLR-normalized data were the most consistent with those from other normalization methods (median F1: 0.989, 0.981, and 0.976; IQR: 0.0220, 0.0235, and 0.0480, respectively). In contrast, SCT, SpaNorm, and RUV-III- were the least consistent and tended to include more unique L-Rs (median F1: 0.873, 0.863, and 0.730; IQR: 0.1411, 0.1218, and 0.0814). Conclusion Normalization influences the strength and composition of inferred CCIs in iST. Log-transformed LS-based and spatially-aware methods predicted stronger L-Rs, while RUV-III-NB, SpaNorm, and SCT identified more unique L-Rs. iST platforms characteristics and tissue context may also affect CCI inference results. Overall, our benchmarking provides guidance for normalization selection in spatial interaction studies, and we suggest combining methods from different methodological categories to enhance the diversity and robustness of detected CCIs.

Abstract: B-cell receptor (BCR) repertoire sequencing enables detailed characterization of antibody responses in infection, vaccination, and autoimmunity. However, comprehensive repertoire analysis typically requires integrating multiple specialized tools with incompatible data formats, limiting accessibility for experimental immunologists. We present absuite, a unified R/Bioconductor package providing an integrated workflow for BCR repertoire analysis. absuite implements ten core analytical functions: V-gene usage profiling, convergent clonotype identification across samples, clonality quantification via Shannon entropy, isotype class-switching analysis, diversity calculations, clonotype similarity networks with community detection, and statistical repertoire comparison. The package accepts data from major repositories (OAS, IMGT) and standard formats (AIRR, CSV), integrating with Bioconductor data structures to enable reproducible, interoperable immune repertoire analysis workflows. We validated absuite using 187,342 antibody sequences from three published COVID-19 cohorts. V-gene analysis correctly identified established COVID-19 signatures (IGHV3-30, IGHV4-34 enrichment). Convergent analysis detected 57 public clonotypes shared across multiple patients, consistent with published findings. Network analysis using Louvain clustering revealed 447 communities with high modularity (0.86), exposing clonotype relationships and somatic variant lineages. Clonality metrics distinguished expanded clonal responses from diverse repertoires, with values ranging from 0.001 to 0.11 across samples. absuite consolidates BCR repertoire analysis into a single Bioconductor-compatible R package, eliminating the need for multiple tools and custom scripting while supporting reproducible immune repertoire workflows. The package is currently being prepared for submission to Bioconductor.

Rare variant burden testing is a powerful approach for identifying genes and pathways enriched for low-frequency, potentially deleterious variants that are often missed by single-variant association analyses. Such approaches are particularly valuable in rare disease and pediatric cancer studies, where genetic risk may be distributed across many variants with modest individual effects. However, burden testing is sensitive to population structure and technical differences across sequencing datasets, and often relies on external control cohorts when internal controls are limited or unavailable. We present burdenBalanceR, an R package that addresses these challenges through two key innovations. First, the package incorporates ancestry estimation using AEon to enable ancestry-matched resampling of cases and controls, providing robust inference of rare variant burden and odds ratios in the presence of population heterogeneity. Second, we implement depth-aware correction methods to account for differences in sequencing coverage between cohorts. By combining ancestry-matched resampling with technical bias correction, this framework improves the reliability of rare variant association analyses. The package supports end-to-end workflows from Variant Effect Predictor (VEP)- annotated VCFs to publication-ready outputs, including variant-level quality control, filtering for rare, predicted deleterious variants in genes or transcripts of interest, and statistical inference using ancestry-matched resampling. We demonstrate its utility using pediatric leukemia cohorts compared with control datasets from the Human Genome Diversity Project and the 1000 Genomes Project, illustrating how ancestry-aware rebalancing mitigates biases arising from heterogeneous ancestral composition and improves comparability across cohorts. The package is available on GitHub and will be submitted to Bioconductor upon completion of suitable vignettes.

DNA methylation array analysis is commonly performed using probe-wise or region-wise classical, frequentist methods. However, many epigenetic mechanistic studies are driven by more targeted biological questions focused on predefined regulatory regions rather than unbiased methylome-wide screening. Examples include Polycomb-associated loci, CpG islands, bivalent promoters, lamina-associated domains, and other chromatin-defined features where coordinated regional behavior is often more biologically meaningful than isolated probe-level significance. For these use cases, a Bayesian framework offers an attractive alternative by providing direct, interpretable probabilities about the direction, magnitude, and uncertainty of methylation change. Here I present BREAD, an R package for Bayesian region-specific DNA methylation inference using array-derived methylation data stored in SummarizedExperiment objects. BREAD is designed for user-defined genomic features and applies hierarchical modeling to borrow strength across CpGs within regions while accommodating covariates and structured experimental designs. Rather than returning only p-values or false discovery rates, BREAD estimates posterior probabilities that each region is hypermethylated or hypomethylated, as well as the probability that methylation differences exceed a user-defined biologically meaningful threshold. This enables outputs that are directly aligned with common biological questions, such as whether aging is associated with coherent methylation gain at selected chromatin-regulatory loci. As a motivating example, I will demonstrate a young-versus-old analysis to assess age-associated methylation shifts across targeted feature sets. This example highlights how region-level Bayesian inference can provide decision-ready answers for biologically coherent loci, including whether a region shows credible hypermethylation, hypomethylation, or insufficient evidence under explicit posterior probability and effect-size criteria. By focusing on predefined regions of interest, BREAD complements existing Bioconductor methylation workflows by addressing a different inferential niche; targeted, uncertainty-aware epigenomic analysis with interpretable feature-level conclusions. BREAD is intended as a flexible framework for methylation studies where researchers care not only about detecting change, but about making direct probabilistic statements on specific regulatory regions. In this way, it extends the Bioconductor ecosystem toward more question-driven, region-resolved DNA methylation inference.

Non-B-form DNA structures, including G-quadruplex, Z-DNA, cruciform, triple helix (H-DNA), are increasingly recognized as important contributors to genome instability. Prior work by Bacolla et al. in 2016 showed that translocation and deletion breakpoints in cancer genomes are associated with putative non-B DNA-forming sequences, providing an important genome-wide view of this relationship. Since then, improved prediction tools for specific non-B conformations, such as Z-DNA Hunter (Petrovič et al., 2025) and G4Hunter (Brázda et al., 2019), together with the growing availability of large cancer sequencing datasets, create an opportunity to revisit this question with higher resolution and broader disease representation. We systematically re-annotated putative non-B-form DNA sequences across the human genome and re-evaluated their colocalization with structural variant loci from a large number of pediatric and adult cancers. We assayed a broad structural variant collection from COSMIC, including TCGA-derived and other curated cancer cohorts. Motivated by the diverse spectrum of fusion events observed in pediatric acute myeloid leukemia (pAML), we hypothesize that unresolved non-B-form secondary DNA structures may increasingly impact genome instability in young patients, who’s genomes have had less time to accumulate large numbers of somatic point mutations, and may therefore be more strongly shaped by structural rearrangements during leukemogenesis. This part of our study focuses on pAML cases from the TARGET initiative. By integrating updated non-B-form annotations with breakpoint- and interval-level structural variant calls, we examined whether specific classes of non-B DNA propensity sequences are enriched near rearrangement-associated regions and fusion breakpoints in pAML. To improve accessibility for other groups, we are organizing this analysis as a reproducible Bioconductor-oriented workflow in R that integrates multiple non-B prediction tools with overlap, colocalization, and enrichment analyses for user-supplied structural variant datasets. This updated framework provides a practical foundation for studying genome instability, structural rearrangement, and developmental context across cancers.

Oncogenic mutations are critical for cancer initiation and maintenance. While this has spurred efforts to inhibit oncoproteins as a strategy for cancer treatment, targetable mutations are found only in a small fraction of cancers. This necessitates broader intervention strategies, such as targeting non-mutated proteins or the molecular circuitry essential to cancer cells. The complexity of cell regulation and malignant transformation create formidable barriers to elucidating disease-critical genetic circuitry. We propose that linking genetic perturbation experiments with statistical modeling can provide inroads to the discovery of functionally relevant gene regulatory network architecture and thus identification of genetic interactions in cancer gene networks, ultimately informing next-generation cancer interventions. The recent development of Perturb-seq, multiplexed CRISPR gene editing coupled with single-cell RNA sequencing, enables gene regulatory network estimation based on thousands of gene perturbations at single cell resolution. However, to be useable for gene regulatory network inference, Perturb-seq data requires careful preprocessing and analysis. We propose a novel approach for Perturb-seq data preprocessing and analysis using control sample matching and a finite mixture modeling approach. Leveraging Perturb-seq data sets ranging in size from less than 100 to nearly 10,000 perturbations, we demonstrate the advantages of our method over other existing methods. Next, we discuss our network modeling methodology, TopNet, which can identify complex non-linear gene interactions while incorporating uncertainty in the underlying gene perturbation data. Finally, we assess the impact of preprocessing and analysis methods on network inference and demonstrate the utility of network modeling to estimate a sparse topological gene network architecture.

Bioconductor's main interface to Gene Ontology is the GO.db package, which leverages AnnotationDbi. These are very bespoke facilities. We consider some of the additional features of ontologies that can be accessed while working with products of the Semantic SQL project at github.com/INCAtools. ontoProc2 supports acquisition and use of any open biological ontology. We will discuss the underpinnings of ontology acquisition and caching, interfacing to the ontologyX suite in CRAN, and operations with Gene Ontology, Cell Ontology, and Experimental Factor Ontology. We will discuss approaches to "retooling" package code to achieve simpler and more comprehensive access to ontological information than is currently provided by AnnotationDbi-based representations.

Spatial biology spans multiple length scales, from intracellular organization to tissue-level architecture. Spatial transcriptomics captures this structure, yet most analyses operate at a single spatial resolution, implicitly assuming that biological organization is scale-consistent. In practice, spatial autocorrelation and co-localization are functions of scale, and conclusions can depend on arbitrary aggregation choices. Here we present Wayfarer, a multiscale framework for spatial -omics that tracks how spatial association metrics evolve across nested spatial aggregations, enabling statistical comparison of multiscale structure across biological conditions. Using Xenium data from lung adenocarcinoma (LUAD), we show that spatial patterns often co-exist at fine and coarse scales and that progression is accompanied by reproducible shifts in scale–response profiles. These include increased fine-scale coherence of ERBB2-high tumor regions and coarse-scale clustering of immune-associated markers that are not apparent at a single resolution. Wayfarer converts spatial aggregation from a confounder into a diagnostic signal and is implemented as an R package to be released through Bioconductor.

Spatial omics technologies provide detailed measurements of cellular phenotypes together with their spatial locations within tissues, enabling new insights into how cellular organization shapes tissue architecture. However, most existing clustering approaches focus on identifying contiguous spatial clusters or clusters defined by a combination of spatial and non-spatial features. As a result, they are limited in their ability to detect repeated spatial patterns, spatially separated tissue regions that share similar underlying phenotype distributions. To address this limitation, we introduce repSpat, a nonparametric statistical inference framework designed to detect repeated spatial patterns in spatial omics data. The framework integrates constrained clustering with a block-permutation inference procedure based on the maximum mean discrepancy (MMD) statistic to formally test for repeated spatial domains. In this workshop, we demonstrate the complete repSpat workflow, including how to convert tabular spatial omics data into AnnData objects, construct spatial constraints, perform constrained hierarchical clustering, apply the repSpat inference framework, relabel detected repeated spatial patterns, and interpret the resulting tissue domains, including tumour microenvironments. We illustrate the workflow using a triple negative breast cancer spatial proteomics dataset, which includes both continuous and multivariate binary markers, as well as simulated examples. These demonstrations highlight the robustness and effectiveness of repSpat in identifying repeated spatial domains in spatial omics datasets.

Computational methods are essential tools for modern microbiome research. Yet, challenges such as lack of standardization, reproducibility, and transparency often limit reliable analysis and interpretation. Bioconductor addresses these challenges through a global, community-driven network that provides a robust, open-source ecosystem of high-quality tools. The presentation highlights how collaborative development supports microbiome research. In this domain, a growing ecosystem of packages builds on shared Bioconductor data structures such as TreeSummarizedExperiment, which extends the widely adopted SummarizedExperiment class. By relying on interoperable classes and consistent design principles, tools developed by different groups can work seamlessly together. For users, this reduces time spent on data wrangling, broadens access to well-established statistical methods, and facilitates the application and benchmarking of approaches developed across different fields. The collaborative nature of the ecosystem ensures that methods and workflows continue to advance through contributions and shared expertise from a diverse community of developers and users. This common foundation enables robust and evidence-based analysis workflows while supporting methodological innovation. It lowers barriers for collaboration between package authors, encourages method reuse, and accelerates cross-disciplinary exchange. As microbiome research progresses toward multi-omics integration, longitudinal studies, and other increasingly complex study designs, interoperable infrastructure supported by an engaged community becomes increasingly critical for scalable and reproducible analyses. Orchestrating Microbiome Analysis (OMA) online book