Keynote
@@ -23,201 +24,42 @@ The full line-up is below the overview — jump to the [contributed talks](#cont
Day 1 · Mon12 October
-
08:00 · Breakfast · Demant
-
09:00
Welcome & State of scverse
Glass Hall
-
09:45 · Keynote
Glass Hall
-
10:30 · Coffee · Demant
-
11:00 · Session 1
Glass Hall
+
08:00 · Breakfast · Demant Hall
+
09:00
Welcome & State of scverse
Glass Hall (Glassalen)
+
09:45 · Keynote
Glass Hall (Glassalen)
+
10:30 · Coffee · Demant Hall
+
11:00 · Session 1
Glass Hall (Glassalen)
12:00 · Lunch · Canteen
-
13:30 · Keynote
Glass Hall
-
14:15 · Session 2
Glass Hall
-
15:00 · Coffee · Demant
-
15:30
Panel discussion
Glass Hall
-
16:30
Poster session 1
Demant
+
13:30 · Keynote
Glass Hall (Glassalen)
+
14:15 · Session 2
Glass Hall (Glassalen)
+
15:00 · Coffee · Demant Hall
+
15:30
Panel discussion
Glass Hall (Glassalen)
+
16:30
Poster session 1
Demant Hall
18:00
Conference dinner & networking
Faculty Club
Day 2 · Tue13 October
-
08:00 · Breakfast · Demant
-
09:00
Opening & housekeeping
Glass Hall
-
09:15 · Keynote
Glass Hall
-
10:00 · Session 3
Glass Hall
-
10:45
Poster session 2
Demant
+
08:00 · Breakfast · Demant Hall
+
09:00
Opening & housekeeping
Glass Hall (Glassalen)
+
09:15 · Keynote
Glass Hall (Glassalen)
+
10:00 · Session 3
Glass Hall (Glassalen)
+
10:45
Poster session 2
Demant Hall
12:00 · Lunch · Canteen
-
13:30 · Keynote
Glass Hall
-
14:15 · Session 4
Glass Hall
-
15:15 · Coffee · Demant
-
15:45 · Workshop
10x Genomics
Glass Hall
-
16:30
Closing remarks
Glass Hall
+
13:30 · Keynote
Glass Hall (Glassalen)
+
14:15 · Session 4
Glass Hall (Glassalen)
+
15:15 · Coffee · Demant Hall
+
15:45 · Workshop
Glass Hall (Glassalen)
+
16:30
Closing remarks
Glass Hall (Glassalen)
18:30
Evening social event (Optional)
Day 3 · Wed14 October
-
08:00 · Breakfast · Demant
-
09:00 · Workshop
Anthropic
Glass Hall
-
10:30 · Coffee · Demant
-
11:00 · Parallel
NVIDIA
S09
-
11:00 · Parallel
Can Ergen
S01
+
08:00 · Breakfast · Demant Hall
+
09:00 · Workshop
Glass Hall (Glassalen)
+
10:30 · Coffee · Demant Hall
+
12:30 · Lunch · Canteen
-
13:30 · Parallel
Matthias Meyer Bender
S01
-
13:30 · Parallel
Stellaromics
S09
-
14:30
Networking & close
Demant
+
+
14:30
Networking & close
Demant Hall
-
-## Contributed talks
-
-Contributed and flash talks run across four themed sessions on Days 1 and 2, selected from the submitted abstracts.
-Full abstracts for every talk will be available in the final programme.
-
-
-
-
- Day 1 · Mon 12 Oct
-
Models, scaling & agents
- 11:00–12:00
-
-
- - 11:00acumenPau Badia i Mompel
- - 11:20SLIMDewei Hu
- - 11:35ScirpyFelix Petschko
- - 11:50scAtlasTb FlashMichaela Mueller
- - 11:55Federated virtual cells FlashArian Amani
-
-
-
-
- Day 1 · Mon 12 Oct
-
Data structures & R/Python interop
- 14:15–15:00
-
-
- - 14:15Unified data modelDaniele Bottazzi
- - 14:35Bioconductor / scverse bridgeArtür Manukyan
- - 14:50Icechunk FlashLuiz Irber
-
-
-
-
- Day 2 · Tue 13 Oct
-
Proteomics
- 10:00–10:45
-
-
- - 10:00AlphaPeptToolsVincenth Brennsteiner
- - 10:20R/Python proteomicsLaurent Gatto
- - 10:35Mantpy FlashMohamed Ghafoor
-
-
-
-
- Day 2 · Tue 13 Oct
-
Spatial omics
- 14:15–15:15
-
-
- - 14:15Squidpy 2.0Tim Treis
- - 14:30Scalable + interactiveArne Defauw
- - 14:45spatialdata.js / MDVPeter Todd
- - 15:00AIH liver FlashFranz Ake
- - 15:05MERFISHEYES FlashIgnatius Jenie
-
-
-
-
-## Keynote talks
-
-### AI for Image-based Systems Biology: From Cryo-Electron Tomography to Virtual Spatial Transcriptomics
-
-**Tingying Peng** · Helmholtz Munich
-
-Recent advances in biological imaging technologies, including cryo-electron tomography and light-sheet microscopy, are enabling the visualization of cellular structures across multiple spatial and temporal scales. However, extracting quantitative insights from increasingly large and complex datasets remains a major challenge. In this talk, I will present our work on developing artificial intelligence methods for quantitative bioimage analysis. I will highlight our recent work on MemBrain v2, an AI framework for automated analysis of membrane organization in cryo-electron tomography that enables large-scale detection of membrane-associated protein complexes directly in situ. MemBrain integrates automated membrane segmentation, geometry-aware particle detection, and downstream analysis of protein organization. I will also briefly discuss additional projects from our group that develop AI methods for microscopy image enhancement and analysis, including tools for illumination correction, such as BaSiCPy, and light-sheet microscopy data processing, such as the Leonardo-toolset.
-
-Finally, I will present our recent work on Phoenix, a generative AI framework for virtual spatial transcriptomics from routine histology. Phoenix integrates multimodal information across tissue morphology, cell states, and gene expression to infer spatially resolved single-cell molecular profiles in situ. Applied across large patient cohorts and disease contexts, Phoenix enables in silico analysis of treatment response, discovery of spatial biomarkers, and prediction of disease-associated tissue organization. This work extends quantitative biological imaging beyond image enhancement and object detection toward multimodal data integration and predictive modeling of molecular tissue states. Together, these approaches aim to enable scalable, quantitative, and predictive analysis of biological imaging data and facilitate new discoveries in structural, cellular, and spatial biology.
-
-### The Rosetta Stone for Human Disease
-
-**Muzlifah Haniffa** · Wellcome Sanger Institute & University of Cambridge
-
-The revolution in single cell genomics, complemented by more recent developments in spatial technologies and artificial intelligence, has changed our understanding of the cellular building blocks of the human body and how our cells form the tissue ecosystems crucial for function. To map the trillions of cells of the body, a large community of researchers from across the globe have assembled under a global consortium, the Human Cell Atlas (HCA). In her talk, Muzlifah Haniffa will discuss the impact of the HCA on modern biomedicine, and then her own work using spatially resolved single-cell genomics and artificial intelligence to decode human development. Her lab has generated the Human Developmental Cell Atlas (HDCA), a unified resource that integrates published and unpublished single-cell/nucleus RNAseq atlases as well as a spatially resolved multimodal cell atlas of 6PCW whole embryos. The HDCA contains ~4.6M cells and ~500 cell types, which were resolved into 94 tissue niches during early development using unsupervised deep learning. This allows description of the development of distributed networks such as the stroma, vasculature and the peripheral nervous system. The HDCA thus provides a holistic window into how human tissue ecosystems are built.
-
-
More keynote talks to be announced.
-
-## Workshops
-
-Hands-on workshops run on Day 2 and Day 3. On Day 3, the opening session is plenary — for all participants — and the two later slots run as parallel tracks, so you can pick one in each.
-
-### From Data to Insight: Hands-On with Sentira Single Cell
-
-**10x Genomics** · Day 2
-
-Join 10x Genomics for an exclusive, hands-on introduction to Sentira Single Cell — 10x Genomics' newly announced autonomous AI agent for Chromium analysis, designed to transform how you navigate high-dimensional omics data.
-
-The exponential growth, scale, and resolution of single-cell and spatial omics have fundamentally transformed our understanding of biology and disease. However, the sheer size of some datasets remains a persistent bottleneck. While modern pipelines process data efficiently, they still demand deep intuition and significant technical acumen to select optimal analytical strategies.
-
-To overcome these barriers, 10x Genomics introduces Sentira Single Cell, a highly scalable framework that democratizes omics analysis through autonomous, LLM-driven workflows. Crucially, Sentira is built upon the trusted ecosystem of core scverse packages. By orchestrating popular tools like Scanpy, AnnData, and Azimuth under the hood, Sentira seamlessly bridges the gap between biological intuition and computational execution. We will walk through how the multi-agent system can decompose a user-specified hypothesis into a bounded execution plan, evaluate intermediate results against explicit success criteria, and adaptively replan to ensure robust, reproducible discoveries.
-
-**What to expect:** a fully immersive, interactive session where you use Sentira Single Cell for complex analytical tasks, with low-latency real-time visualization and interactive downstream analysis in a streamlined web interface. Participants will explore how Sentira translates natural language into execution, self-evaluates and adapts, and accelerates high-plex discovery.
-
-**Bring your own data:** following a brief overview of the platform's architecture, attendees are highly encouraged to bring their own single-cell data (e.g. scRNA-seq in `.h5` format) to run live through Sentira during the hands-on segment. For those without data on hand, 10x Genomics will provide benchmark datasets to explore and analyze.
-
-**Prepare in advance.** To take part in the hands-on segment, please complete both of these before the session:
-
-- Sign up for a 10x Cloud Analysis account — [sign up or sign in](https://www.10xgenomics.com/products/cloud-analysis).
-- Request early access for Sentira — [10xgenomics.com/software/sentira](https://www.10xgenomics.com/software/sentira).
-
-### Hands-on single-cell analysis with Claude Science
-
-**Anthropic**
-
-Claude Science is Anthropic's new AI workbench for scientific research, currently in public beta. It runs on your own laptop or server with the standard Claude models, and adds what a researcher needs around them, including persistent Python and R sessions that you can drive with varying levels of coding experience, connections to more than 60 public scientific databases, and analysis specialists for single-cell, genomics, proteomics and other domains. Everything it produces is saved with the code, environment and reasoning that generated it, so a result can be reproduced and audited by you or anyone in your lab long after the session ends.
-
-In this workshop we will go over a real-life application in the single-cell space. Starting from public data or your own, we will showcase Claude Science with an end-to-end analysis of single-cell data that leads to biological insight, driven by conversation and built on scverse tools. Along the way we will look at how Claude plans an analysis, how it writes and runs real code you can read and keep, how it checks its own work, and where a scientist still needs to step in and steer. The session is aimed at people with light coding experience and a strong interest in getting answers out of single-cell experiments. If you can describe the biological question, we will show you how far Claude Science can take you toward the analysis, and how to stay in control of what it did and why.
-
-### scviva-tools: From Niche Embeddings to Gene Modules
-
-**Can Ergen** · scverse
-
-scviva-tools is a consolidated, scverse-native spatial transcriptomics toolkit built on scvi-tools, unifying probabilistic models and downstream analysis under one pip-installable API. The model layer includes ResolVI (denoising and segmentation-error correction), DestVI (multi-resolution cell-type deconvolution), scVIVA (niche-aware representation learning), DiagVI, gimVI, Stereoscope, and Tangram, with more models added as the community contributes. The downstream layer consumes these outputs, or raw annotated spatial data directly, to answer targeted biological questions: Harreman infers metabolic exchange and cell-cell communication, CSDE recovers de-biased differential expression, VIVS identifies genes conditionally dependent on an external response, and VISION scores gene signatures for spatial autocorrelation. Every downstream tool reads directly from a model's output written into the datamodule's .obs/.layers/latent space, so results compose without custom glue code. In this one hour workshop we focus on one complete pipeline rather than a full tour of the toolkit. Participants will run scVIVA on a spatial dataset to learn niche-aware representations, then feed that embedding directly into Harreman to identify gene modules underlying cell-cell communication within those niches with no custom glue code required, since Harreman reads straight from scVIVA's output in the shared datamodule. Along the way we'll show how the rest of scviva-tools extends from this same foundation, so participants leave with a working local install, a runnable scVIVA to Harreman notebook they can adapt to their own data, and a clear sense of where the other models and tools fit in.
-
-### SegTraQ: A toolbox to assess segmentation and transcript assignment quality in spatial transcriptomics data
-
-**Matthias Meyer-Bender**
-
-Cell segmentation in spatial transcriptomics data is hampered by several technical factors:
-
-- Cells may be sectioned without their nuclei, leading to undersegmentation.
-- Overlapping cells in the z-plane can appear as one cell in 2D projections, leading to mixed expression profiles.
-- Transcript diffusion can occur during tissue processing, contaminating neighboring cells.
-
-To address these limitations, transcript-informed segmentation methods leverage spatial co-expression patterns of transcripts. However, evaluating the quality of the segmentation is difficult due to the high-dimensional and sparse nature of the data and the lack of manually curated datasets.
-
-We introduce SegTraQ, a Python-based framework for segmentation and transcript assignment quality control in spatial transcriptomics data. SegTraQ computes quantitative metrics designed to highlight regions or samples with poorly segmented cells, and guide the choice of appropriate segmentation methods. The package is composed of six modules, each of which addresses a different aspect of segmentation quality, such as undersegmentation, mutually exclusive coexpression rate, or unresolved overlap between cells in 3D. Based on the spatialdata format, SegTraQ integrates seamlessly into the scverse ecosystem.
-
-In this workshop, we demonstrate how these metrics can be used to compare different segmentations, both between samples and between algorithms.
-
-The session opens with a 15-minute introduction to the topic, after which participants work through a notebook hands-on with the data and packages.
-
-**Prepare in advance.** The workshop materials are on [GitHub](https://github.com/MeyerBender/segtraq_workshop). Please download the dataset (~600 MB) [from here](https://oc.embl.de/index.php/s/bBj36ET5S3v5VMW) before the session — depending on network capacity on the day, there may also be time to download it on-site.
-
-### GPU-Accelerated Single-Cell Genomics: Tools, Workflows, and Spatial Analysis
-
-**NVIDIA** · Trainers: Severin Dicks, Lukas Heumos, Sara Jimenez · Support: Heidi Shin
-
-**Objective:** participants will understand how to GPU-accelerate single-cell and spatial workloads using the scverse ecosystem through a familiar single-cell workflow run using RAPIDS-singlecell.
-
-Single-cell genomics datasets have grown exponentially, from thousands of cells per sample to millions in a single experimental design. This shift moves the field from method development to method acceleration, enabling researchers to answer existing biological questions at previously impossible scale. NVIDIA develops software libraries and open models to support this acceleration, including CUDA-X, Parabricks, and BioNemo. Together with scverse, we introduce rapids-singlecell, a GPU-accelerated implementation of foundational single-cell tools like scanpy, squidpy, pertpy, and decoupler. Through scverse-backends, users can seamlessly switch between CPU and GPU analysis based on computational needs and data scale. This workshop combines short lectures with hands-on examples. We demonstrate a real-world case study using 10x Genomics' Atera technology — a high-throughput image-based platform generating whole-transcriptome data. We walk through the complete analysis pipeline step-by-step, highlighting new functionalities for spatial niche detection and evaluation.
-
-### 3D spatial transcriptomics with Stellaromics
-
-**Stellaromics**
-
-A hands-on tutorial working with 3D spatial transcriptomics data from the Stellaromics commercial platform.
-
-
Full abstract to follow.
-
-## Posters
-
-Poster boards are **160 × 120 cm (W × H)**, in landscape orientation.
-Please keep your poster to a maximum of **155 × 115 cm (W × H)** so it fits the board.
-In the A series, **A0 landscape (118.9 × 84.1 cm)** fits comfortably.
diff --git a/content/conference2026/sponsors.md b/content/conference2026/sponsors.md
index 4503f156..00f658db 100644
--- a/content/conference2026/sponsors.md
+++ b/content/conference2026/sponsors.md
@@ -2,7 +2,7 @@
title = "Sponsors"
url = "/conference2026/sponsors"
type = "conference2026"
-weight = 6
+weight = 8
+++
## Our sponsors
diff --git a/content/conference2026/talks.md b/content/conference2026/talks.md
new file mode 100644
index 00000000..acab5285
--- /dev/null
+++ b/content/conference2026/talks.md
@@ -0,0 +1,93 @@
++++
+title = "Talks"
+url = "/conference2026/talks"
+type = "conference2026"
+weight = 4
++++
+
+Contributed talks, keynote abstracts, and poster guidelines for scverse conference 2026.
+See the [schedule](/conference2026/schedule/) for the full day-by-day programme.
+
+## Contributed talks
+
+Contributed and flash talks run across four themed sessions on Days 1 and 2, selected from the submitted abstracts.
+Full abstracts for every talk will be available in the final programme.
+
+
+
+
+ Day 1 · Mon 12 Oct
+
Models, scaling & agents
+ 11:00–12:00
+
+
+ - 11:00acumenPau Badia i Mompel
+ - 11:20SLIMDewei Hu
+ - 11:35ScirpyFelix Petschko
+ - 11:50scAtlasTb FlashMichaela Mueller
+ - 11:55Federated virtual cells FlashArian Amani
+
+
+
+
+ Day 1 · Mon 12 Oct
+
Data structures & R/Python interop
+ 14:15–15:00
+
+
+ - 14:15Unified data modelDaniele Bottazzi
+ - 14:35Bioconductor / scverse bridgeArtür Manukyan
+ - 14:50Icechunk FlashLuiz Irber
+
+
+
+
+ Day 2 · Tue 13 Oct
+
Proteomics
+ 10:00–10:45
+
+
+ - 10:00AlphaPeptToolsVincenth Brennsteiner
+ - 10:20R/Python proteomicsLaurent Gatto
+ - 10:35Mantpy FlashMohamed Ghafoor
+
+
+
+
+ Day 2 · Tue 13 Oct
+
Spatial omics
+ 14:15–15:15
+
+
+ - 14:15Squidpy 2.0Tim Treis
+ - 14:30Scalable + interactiveArne Defauw
+ - 14:45spatialdata.js / MDVPeter Todd
+ - 15:00AIH liver FlashFranz Ake
+ - 15:05MERFISHEYES FlashIgnatius Jenie
+
+
+
+
+## Keynote talks
+
+### AI for Image-based Systems Biology: From Cryo-Electron Tomography to Virtual Spatial Transcriptomics
+
+**Tingying Peng** · Helmholtz Munich
+
+Recent advances in biological imaging technologies, including cryo-electron tomography and light-sheet microscopy, are enabling the visualization of cellular structures across multiple spatial and temporal scales. However, extracting quantitative insights from increasingly large and complex datasets remains a major challenge. In this talk, I will present our work on developing artificial intelligence methods for quantitative bioimage analysis. I will highlight our recent work on MemBrain v2, an AI framework for automated analysis of membrane organization in cryo-electron tomography that enables large-scale detection of membrane-associated protein complexes directly in situ. MemBrain integrates automated membrane segmentation, geometry-aware particle detection, and downstream analysis of protein organization. I will also briefly discuss additional projects from our group that develop AI methods for microscopy image enhancement and analysis, including tools for illumination correction, such as BaSiCPy, and light-sheet microscopy data processing, such as the Leonardo-toolset.
+
+Finally, I will present our recent work on Phoenix, a generative AI framework for virtual spatial transcriptomics from routine histology. Phoenix integrates multimodal information across tissue morphology, cell states, and gene expression to infer spatially resolved single-cell molecular profiles in situ. Applied across large patient cohorts and disease contexts, Phoenix enables in silico analysis of treatment response, discovery of spatial biomarkers, and prediction of disease-associated tissue organization. This work extends quantitative biological imaging beyond image enhancement and object detection toward multimodal data integration and predictive modeling of molecular tissue states. Together, these approaches aim to enable scalable, quantitative, and predictive analysis of biological imaging data and facilitate new discoveries in structural, cellular, and spatial biology.
+
+### The Rosetta Stone for Human Disease
+
+**Muzlifah Haniffa** · Wellcome Sanger Institute & University of Cambridge
+
+The revolution in single cell genomics, complemented by more recent developments in spatial technologies and artificial intelligence, has changed our understanding of the cellular building blocks of the human body and how our cells form the tissue ecosystems crucial for function. To map the trillions of cells of the body, a large community of researchers from across the globe have assembled under a global consortium, the Human Cell Atlas (HCA). In her talk, Muzlifah Haniffa will discuss the impact of the HCA on modern biomedicine, and then her own work using spatially resolved single-cell genomics and artificial intelligence to decode human development. Her lab has generated the Human Developmental Cell Atlas (HDCA), a unified resource that integrates published and unpublished single-cell/nucleus RNAseq atlases as well as a spatially resolved multimodal cell atlas of 6PCW whole embryos. The HDCA contains ~4.6M cells and ~500 cell types, which were resolved into 94 tissue niches during early development using unsupervised deep learning. This allows description of the development of distributed networks such as the stroma, vasculature and the peripheral nervous system. The HDCA thus provides a holistic window into how human tissue ecosystems are built.
+
+
More keynote talks to be announced.
+
+## Posters
+
+Poster boards are **160 × 120 cm (W × H)**, in landscape orientation.
+Please keep your poster to a maximum of **155 × 115 cm (W × H)** so it fits the board.
+In the A series, **A0 landscape (118.9 × 84.1 cm)** fits comfortably.
diff --git a/content/conference2026/travel.md b/content/conference2026/travel.md
index a1c761cd..8eba2318 100644
--- a/content/conference2026/travel.md
+++ b/content/conference2026/travel.md
@@ -2,7 +2,7 @@
title = "Venue & travel"
url = "/conference2026/travel"
type = "conference2026"
-weight = 7
+weight = 9
+++
scverse conference 2026 · DTU Lyngby Campus, Kongens Lyngby · 12–14 October 2026
diff --git a/content/conference2026/workshops.md b/content/conference2026/workshops.md
new file mode 100644
index 00000000..0ebaf106
--- /dev/null
+++ b/content/conference2026/workshops.md
@@ -0,0 +1,79 @@
++++
+title = "Workshops"
+url = "/conference2026/workshops"
+type = "conference2026"
+weight = 5
++++
+
+## Workshops
+
+Hands-on workshops run on Day 2 and Day 3. On Day 3, the opening session is plenary — for all participants — and the two later slots run as parallel tracks, so you can pick one in each. See the [schedule](/conference2026/schedule/) for timings and rooms.
+
+### From Data to Insight: Hands-On with Sentira Single Cell
+
+**10x Genomics** · Day 2
+
+Join 10x Genomics for an exclusive, hands-on introduction to Sentira Single Cell — 10x Genomics' newly announced autonomous AI agent for Chromium analysis, designed to transform how you navigate high-dimensional omics data.
+
+The exponential growth, scale, and resolution of single-cell and spatial omics have fundamentally transformed our understanding of biology and disease. However, the sheer size of some datasets remains a persistent bottleneck. While modern pipelines process data efficiently, they still demand deep intuition and significant technical acumen to select optimal analytical strategies.
+
+To overcome these barriers, 10x Genomics introduces Sentira Single Cell, a highly scalable framework that democratizes omics analysis through autonomous, LLM-driven workflows. Crucially, Sentira is built upon the trusted ecosystem of core scverse packages. By orchestrating popular tools like Scanpy, AnnData, and Azimuth under the hood, Sentira seamlessly bridges the gap between biological intuition and computational execution. We will walk through how the multi-agent system can decompose a user-specified hypothesis into a bounded execution plan, evaluate intermediate results against explicit success criteria, and adaptively replan to ensure robust, reproducible discoveries.
+
+**What to expect:** a fully immersive, interactive session where you use Sentira Single Cell for complex analytical tasks, with low-latency real-time visualization and interactive downstream analysis in a streamlined web interface. Participants will explore how Sentira translates natural language into execution, self-evaluates and adapts, and accelerates high-plex discovery.
+
+**Bring your own data:** following a brief overview of the platform's architecture, attendees are highly encouraged to bring their own single-cell data (e.g. scRNA-seq in `.h5` format) to run live through Sentira during the hands-on segment. For those without data on hand, 10x Genomics will provide benchmark datasets to explore and analyze.
+
+**Prepare in advance.** To take part in the hands-on segment, please complete both of these before the session:
+
+- Sign up for a 10x Cloud Analysis account — [sign up or sign in](https://www.10xgenomics.com/products/cloud-analysis).
+- Request early access for Sentira — [10xgenomics.com/software/sentira](https://www.10xgenomics.com/software/sentira).
+
+### Hands-on single-cell analysis with Claude Science
+
+**Anthropic**
+
+Claude Science is Anthropic's new AI workbench for scientific research, currently in public beta. It runs on your own laptop or server with the standard Claude models, and adds what a researcher needs around them, including persistent Python and R sessions that you can drive with varying levels of coding experience, connections to more than 60 public scientific databases, and analysis specialists for single-cell, genomics, proteomics and other domains. Everything it produces is saved with the code, environment and reasoning that generated it, so a result can be reproduced and audited by you or anyone in your lab long after the session ends.
+
+In this workshop we will go over a real-life application in the single-cell space. Starting from public data or your own, we will showcase Claude Science with an end-to-end analysis of single-cell data that leads to biological insight, driven by conversation and built on scverse tools. Along the way we will look at how Claude plans an analysis, how it writes and runs real code you can read and keep, how it checks its own work, and where a scientist still needs to step in and steer. The session is aimed at people with light coding experience and a strong interest in getting answers out of single-cell experiments. If you can describe the biological question, we will show you how far Claude Science can take you toward the analysis, and how to stay in control of what it did and why.
+
+### scviva-tools: From Niche Embeddings to Gene Modules
+
+**Can Ergen** · scverse
+
+scviva-tools is a consolidated, scverse-native spatial transcriptomics toolkit built on scvi-tools, unifying probabilistic models and downstream analysis under one pip-installable API. The model layer includes ResolVI (denoising and segmentation-error correction), DestVI (multi-resolution cell-type deconvolution), scVIVA (niche-aware representation learning), DiagVI, gimVI, Stereoscope, and Tangram, with more models added as the community contributes. The downstream layer consumes these outputs, or raw annotated spatial data directly, to answer targeted biological questions: Harreman infers metabolic exchange and cell-cell communication, CSDE recovers de-biased differential expression, VIVS identifies genes conditionally dependent on an external response, and VISION scores gene signatures for spatial autocorrelation. Every downstream tool reads directly from a model's output written into the datamodule's .obs/.layers/latent space, so results compose without custom glue code. In this one hour workshop we focus on one complete pipeline rather than a full tour of the toolkit. Participants will run scVIVA on a spatial dataset to learn niche-aware representations, then feed that embedding directly into Harreman to identify gene modules underlying cell-cell communication within those niches with no custom glue code required, since Harreman reads straight from scVIVA's output in the shared datamodule. Along the way we'll show how the rest of scviva-tools extends from this same foundation, so participants leave with a working local install, a runnable scVIVA to Harreman notebook they can adapt to their own data, and a clear sense of where the other models and tools fit in.
+
+### SegTraQ: A toolbox to assess segmentation and transcript assignment quality in spatial transcriptomics data
+
+**Matthias Meyer-Bender**
+
+Cell segmentation in spatial transcriptomics data is hampered by several technical factors:
+
+- Cells may be sectioned without their nuclei, leading to undersegmentation.
+- Overlapping cells in the z-plane can appear as one cell in 2D projections, leading to mixed expression profiles.
+- Transcript diffusion can occur during tissue processing, contaminating neighboring cells.
+
+To address these limitations, transcript-informed segmentation methods leverage spatial co-expression patterns of transcripts. However, evaluating the quality of the segmentation is difficult due to the high-dimensional and sparse nature of the data and the lack of manually curated datasets.
+
+We introduce SegTraQ, a Python-based framework for segmentation and transcript assignment quality control in spatial transcriptomics data. SegTraQ computes quantitative metrics designed to highlight regions or samples with poorly segmented cells, and guide the choice of appropriate segmentation methods. The package is composed of six modules, each of which addresses a different aspect of segmentation quality, such as undersegmentation, mutually exclusive coexpression rate, or unresolved overlap between cells in 3D. Based on the spatialdata format, SegTraQ integrates seamlessly into the scverse ecosystem.
+
+In this workshop, we demonstrate how these metrics can be used to compare different segmentations, both between samples and between algorithms.
+
+The session opens with a 15-minute introduction to the topic, after which participants work through a notebook hands-on with the data and packages.
+
+**Prepare in advance.** The workshop materials are on [GitHub](https://github.com/MeyerBender/segtraq_workshop). Please download the dataset (~600 MB) [from here](https://oc.embl.de/index.php/s/bBj36ET5S3v5VMW) before the session — depending on network capacity on the day, there may also be time to download it on-site.
+
+### GPU-Accelerated Single-Cell Genomics: Tools, Workflows, and Spatial Analysis
+
+**NVIDIA** · Trainers: Severin Dicks, Lukas Heumos, Sara Jimenez · Support: Heidi Shin
+
+**Objective:** participants will understand how to GPU-accelerate single-cell and spatial workloads using the scverse ecosystem through a familiar single-cell workflow run using RAPIDS-singlecell.
+
+Single-cell genomics datasets have grown exponentially, from thousands of cells per sample to millions in a single experimental design. This shift moves the field from method development to method acceleration, enabling researchers to answer existing biological questions at previously impossible scale. NVIDIA develops software libraries and open models to support this acceleration, including CUDA-X, Parabricks, and BioNemo. Together with scverse, we introduce rapids-singlecell, a GPU-accelerated implementation of foundational single-cell tools like scanpy, squidpy, pertpy, and decoupler. Through scverse-backends, users can seamlessly switch between CPU and GPU analysis based on computational needs and data scale. This workshop combines short lectures with hands-on examples. We demonstrate a real-world case study using 10x Genomics' Atera technology — a high-throughput image-based platform generating whole-transcriptome data. We walk through the complete analysis pipeline step-by-step, highlighting new functionalities for spatial niche detection and evaluation.
+
+### 3D spatial transcriptomics with Stellaromics
+
+**Stellaromics**
+
+A hands-on tutorial working with 3D spatial transcriptomics data from the Stellaromics commercial platform.
+
+
Full abstract to follow.