|
| 1 | ++++ |
| 2 | +title = "Four new scverse core packages" |
| 3 | +date = 2025-07-01T00:00:05+01:00 |
| 4 | +description = "scverse expands with new core packages." |
| 5 | +author = "Lukas Heumos" |
| 6 | +draft = false |
| 7 | ++++ |
| 8 | + |
| 9 | +# Four new core packages in scverse |
| 10 | + |
| 11 | +We're happy to announce that four new packages have joined the scverse core ecosystem: [SnapATAC2](https://github.com/scverse/snapatac2), [rapids-singlecell](https://github.com/scverse/rapids_singlecell), [pertpy](https://github.com/scverse/pertpy), and [decoupler](https://github.com/scverse/decoupler). |
| 12 | +This broadens the scope of scverse beyond our so far supported modalities and brings in new functionality for epigenomics, perturbation screens, GPU acceleration, and functional inference. |
| 13 | + |
| 14 | +Single-cell analysis is evolving rapidly, with new experimental modalities and larger datasets becoming the norm. |
| 15 | +To keep up with this, we’re expanding scverse to support more domains, more data types, and more scalable computing backends. |
| 16 | +By bringing these mature, well-maintained tools into the core, we aim to provide a consistent and interoperable foundation across all stages of single-cell analysis. |
| 17 | + |
| 18 | +We're also welcoming the following lead maintainers to the scverse core developer team: |
| 19 | +- **Kai Zhang** for **snapatac2** |
| 20 | +- **Pau Badia i Mompel** for **decoupler** |
| 21 | + |
| 22 | +## SnapATAC2 |
| 23 | + |
| 24 | +[SnapATAC2](https://github.com/scverse/snapatac2) enables fast, scalable analysis of single-cell ATAC-seq and related epigenomic data. |
| 25 | +Built in Rust with a Python front end, it handles millions of cells efficiently. |
| 26 | +The package offers preprocessing, dimensionality reduction, clustering, and visualization methods. |
| 27 | +All outputs are stored in AnnData and integrate seamlessly with scanpy and other scverse frameworks. |
| 28 | + |
| 29 | +<img src="/img/blog/snapatac2_overview.png" style="max-width: 100%;" alt="SnapATAC2 overview" /> |
| 30 | + |
| 31 | +## rapids-singlecell |
| 32 | + |
| 33 | +[rapids-singlecell](https://github.com/scverse/rapids_singlecell) accelerates the full single-cell analysis pipeline through GPU acceleration with CuPy and NVIDIA RAPIDS. |
| 34 | +Core steps—including PCA, neighborhood graph construction, and clustering—are executed on GPU using cuML, cuGraph, and custom CUDA/CuPy kernels for peak performance. |
| 35 | + |
| 36 | +RSC integrates directly with AnnData and offers near drop-in replacements not only for scanpy, but also for selected functions from decoupler and squidpy. |
| 37 | +By preserving familiar APIs and data structures, it enables seamless GPU acceleration of existing workflows—scaling to millions of cells without the computational bottlenecks of CPU-based analysis. |
| 38 | + |
| 39 | +<img src="/img/blog/rsc_overview.png" style="max-width: 100%;" alt="RSC overview" /> |
| 40 | + |
| 41 | +For more details, we refer to a recent blog post by NVIDIA: [Driving Toward Billion-Cell Analysis and Biological Breakthroughs with RAPIDS-singlecell](https://developer.nvidia.com/blog/driving-toward-billion-cell-analysis-and-biological-breakthroughs-with-rapids-singlecell) |
| 42 | + |
| 43 | +## pertpy |
| 44 | + |
| 45 | +[pertpy](https://github.com/scverse/pertpy) focuses on single-cell perturbation screens, including CRISPR and compound treatments. |
| 46 | +It supports differential analysis, signature scoring, and dose-response modeling. |
| 47 | +Metadata handling and visualization are tailored for perturbation-specific use cases. |
| 48 | +Built on AnnData and scverse libraries, pertpy fits smoothly into existing pipelines. |
| 49 | + |
| 50 | +<img src="https://github.com/user-attachments/assets/d2e32d69-b767-4be3-a938-77a9dce45d3f" alt="pertpy overview" style="max-width: 100%;" /> |
| 51 | + |
| 52 | +## decoupler |
| 53 | + |
| 54 | +[decoupler](https://github.com/scverse/decoupler) enables the inference of enrichment scores from omics data using prior knowledge. |
| 55 | +It maps omics profiles to annotated biological sets, such as transcription factors, pathways, or kinases, using methods like GSEA, GSVA, and linear models. |
| 56 | +Designed for bulk, single-cell and spatial data, decoupler works directly with our scverse core data structures. |
| 57 | + |
| 58 | +<img src="/img/blog/decoupler_overview.png" style="max-width: 100%;" alt="decoupler overview" /> |
| 59 | + |
| 60 | +## What this means |
| 61 | + |
| 62 | +We're continuing to support a modular but coherent ecosystem where high-quality tools can interoperate. |
| 63 | +These additions bring coverage of new data types and analysis goals while staying within the same technical foundations. |
| 64 | +All packages use our scverse core data structures, follow shared conventions, and benefit from a growing set of shared infrastructure and community practices. |
| 65 | +We aim to keep development decentralized and open while improving alignment across projects. |
| 66 | + |
| 67 | +## Get involved |
| 68 | + |
| 69 | +If you're building something new, we'd love to have your work be a part of the [scverse ecosystem](https://github.com/scverse/ecosystem-packages). |
| 70 | +If you're our tools, please share feedback or ideas via the respective issue trackers. |
| 71 | +The best places to start are [scverse.org](https://scverse.org), [github.com/scverse](https://github.com/scverse), and our [scverse zulip chat](https://scverse.zulipchat.com/). |
| 72 | +We're always looking for more contributors to our packages but especially also for community related work. |
| 73 | +Please reach out! |
| 74 | + |
| 75 | +## Thank you |
| 76 | + |
| 77 | +We’re grateful to the maintainers of these packages for their work and commitment to open, reusable tools. |
| 78 | +Their contributions help make the scverse ecosystem more useful, inclusive, and sustainable. |
| 79 | +We’re also thankful to the community for using, testing, and contributing to these tools — your feedback drives everything we do. |
| 80 | + |
| 81 | +*— The scverse core team* |
0 commit comments