KO | EN

GitLite

Minimalist Open Source Repository Search Engine

gitlite — search

#dsh #dsh-plugin #developer-tools #ai-agent #codex #typescript #cli #deepseek-harness #claude-code #ai-agents #mcp #rust

scMultiBench

★ 26 Open GitHub ↗

Multi-task benchmarking of single-cell multimodal omics integration methods

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

<h1 style="display: flex; align-items: center; font-size: 2em;">
  <img src="figure/logo.png" alt="Logo" style="width: 50px; height: 50px; margin-right: 10px;">
  scMultiBench
</h1>

Multi-task benchmarking of single-cell multimodal omics integration methods

Single-cell multimodal omics technologies have empowered the profiling of complex biological systems at a resolution and scale that were previously unattainable. These biotechnologies have propelled the fast-paced innovation and development of data integration methods, leading to a critical need for their systematic categorisation, evaluation, and benchmark. Navigating and selecting the most pertinent integration approach poses a significant challenge, contingent upon the tasks relevant to the study goals and the combination of modalities and batches present in the data at hand. Understanding how well each method performs multiple tasks, including dimension reduction, batch correction, cell type classification and clustering, imputation, feature selection, and spatial registration, and at which combinations will help guide this decision. This study aims to develop a much-needed guideline on choosing the most appropriate method for single-cell multimodal omics data analysis through a systematic categorisation and comprehensive benchmarking of current methods.

## Quick Start

Clone the repository and create the lightweight evaluation environment:

```bash
git clone https://github.com/PYangLab/scMultiBench.git
cd scMultiBench
mamba env create -f envs/evaluation-python.yml
mamba activate scmultibench-eval
```

Validate the bundled classification demo data:

```bash
python scripts/validate_inputs.py classification \
  --reference data/classification/demo_data/data1.h5 \
  --query data/classification/demo_data/data2.h5 \
  --reference-labels data/classification/demo_data/cty1.csv \
  --query-labels data/classification/demo_data/cty2.csv
```

Run the bundled MLP classification example:

```bash
python scripts/run_metric.py mlp-classification \
  --reference data/classification/demo_data/data1.h5 \
  --query data/classification/demo_data/data2.h5 \
  --reference-labels data/classification/demo_data/cty1.csv \
  --query-labels data/classification/demo_data/cty2.csv \
  --out results/classification/demo \
  --epochs 10
```

Run the embedding-based scIB-style metrics on the bundled dimension-reduction/batch-correction demo:

```bash
python scripts/run_metric.py scib-embedding \
  --embedding "data/dr&bc/embedding/embedding.h5" \
  --labels "data/dr&bc/embedding/cty1.csv" "data/dr&bc/embedding/cty2.csv" "data/dr&bc/embedding/cty3.csv" \
  --cluster data/clustering/embedding/sinfonia_clustering.h5 \
  --batch-cluster data/clustering/embedding/sinfonia_clustering_batch.h5 \
  --out results/scib_metrics/demo
```

The repository contains many method-specific scripts with incompatible dependency stacks. Use the lightweight environments in `envs/` for validation and core evaluation, then create method-specific environments from the original method documentation when running a full benchmark.

## Repository Layout

- `data/`: small demo inputs and demo outputs.
- `evaluation_pipelines/`: metric and downstream evaluation scripts.
- `tools_scripts/`: method-specific integration scripts.
- `figure/` and `figure_script/`: benchmark figures and plotting scripts.
- `metadata/methods.yaml`: method/task/script registry.
- `scripts/`: utility commands for validation and standard runner dispatch.
- `envs/`: reproducible conda/mamba environment templates.
- `docs/`: input/output contracts and usage notes.

## Integration Tools

In this benchmark, we evaluated 40 integration methods across the four data integration categories on 64 real datasets and 22 simulated datasets on a Ubuntu system with  RTX3090 GPU. In particular, we include 18 vertical integration methods, 14 diagonal integration tools, 12 mosaic integration tools, and 15 cross integration tools. The installation environment is set up according to the respective tutorials. Tools that are compared include:

Vertical Integration (Dimension Reduction and Clustering):
- [totalVI](https://docs.scvi-tools.org/en/stable/tutorials/notebooks/multimodal/totalVI.html) v1.1.2
- [sciPENN](https://github.com/jlakkis/sciPENN) v1.0.0
- [Concerto](https://github.com/melobio/Concerto-reproducibility) Github Version: ab1fc7f
- [scMSI](https://github.com/ChengmingZhang-CAS/scMSI-master) Github Version: dffcbb2
- [Matilda](https://github.com/PYangLab/Matilda) Github Version: 7d71480
- [MOFA+](https://biofam.github.io/MOFA2/) v1.6.0
- [Multigrate](https://multigrate.readthedocs.io/en/latest/index.html) v0.0.2
- [UINMF](http://htmlpreview.github.io/?https://github.com/welch-lab/liger/blob/master/vignettes/UINMF_vignette.html) v2.0.1
- [scMoMaT](https://github.com/PeterZZQ/scMoMaT) v0.2.2
- [Seurat_WNN](https://satijalab.org/seurat/articles/weighted_nearest_neighbor_analysis) v5.0.2
- [scMM](https://github.com/kodaim1115/scMM) Github Version: c5c8579
- [scMDC](https://github.com/xianglin226/scMDC/tree/v1.0.0) Github Version: 43b0c3a
- [moETM](https://github.com/manqizhou/moETM) Github Version: c2eaa97
- [VIMCCA](https://github.com/jhu99/scbean) Github Version: 0.5.6
- [iPOLNG](https://github.com/cuhklinlab/iPoLNG) v0.0.2
- [MIRA](https://github.com/cistrome/MIRA) v2.1.0
- [UnitedNet](https://github.com/LiuLab-Bioelectronics-Harvard/UnitedNet) Github Version: 3689da8
- [scMVP](https://github.com/bm2-lab/scMVP) Github Version: fc61e4d

Vertical Integration (Feature Selection):
- [scMoMaT](https://github.com/PeterZZQ/scMoMaT) v0.2.2
- [Matilda](https://github.com/PYangLab/Matilda) Github Version: 7d71480
- [MOFA+](https://biofam.github.io/MOFA2/) v1.6.0
  
Diagonal Integration (Dimension Reduction, Batch Correction, Clustering, Classification):
- [scBridge](https://github.com/XLearning-SCU/scBridge) Github Version: ff17561
- [Portal](https://github.com/YangLabHKUST/Portal) v1.0.2
- [SCALEX](https://github.com/jsxlei/SCALEX) v1.0.2
- [VIPCCA](https://github.com/jhu99/vipcca) v0.2.7
- [Seurat v3](https://satijalab.org/seurat/articles/seurat5_atacseq_integration_vignette) v5.0.2
- [MultiMAP](https://github.com/Teichlab/MultiMAP) Github Version: 681e608
- [Seurat v5](https://satijalab.org/seurat/articles/seurat5_integration_bridge) v5.0.2
- [sciCAN](https://github.com/rpmccordlab/sciCAN) Github Version: ad71bba 
- [Conos](https://github.com/kharchenkolab/conos) v1.4.6
- [iNMF](https://github.com/welch-lab/liger) v2.0.1
- [online iNMF](http://htmlpreview.github.io/?https://github.com/welch-lab/liger/blob/master/vignettes/UINMF_vignette.html) v2.0.1
- [scJoint](https://github.com/SydneyBioX/scJoint)  Github Version: cbbfa5d
- [GLUE](https://github.com/gao-lab/GLUE) v0.3.2
- [uniPort](https://uniport.readthedocs.io/en/latest/index.html) v1.2.2

Mosaic Integration (Dimension Reduction, Batch Correction, Clustering, Classification):
- [MultiVI](https://docs.scvi-tools.org/en/stable/tutorials/notebooks/multimodal/MultiVI_tutorial.html) v1.1.2
- [scMoMaT](https://github.com/PeterZZQ/scMoMaT) v0.2.2
- [StabMap](https://github.com/MarioniLab/StabMap) v0.1.8
- [Cobolt](https://github.com/epurdom/cobolt) v1.0.1
- [UINMF](http://htmlpreview.github.io/?https://github.com/welch-lab/liger/blob/master/vignettes/UINMF_vignette.html) v2.0.1
- [Multigrate](https://multigrate.readthedocs.io/en/latest/index.html) v0.0.2
- [SMILE](https://github.com/rpmccordlab/SMILE) Github Version: a2e2ca6

Mosaic Integration (Imputation):
- [scMM](https://github.com/kodaim1115/scMM) Github Version: c5c8579
- [moETM](https://github.com/manqizhou/moETM) Github Version: ad89fe2
- [UnitedNet](https://github.com/LiuLab-Bioelectronics-Harvard/UnitedNet) Github Version: 3689da8
- [totalVI](https://docs.scvi-tools.org/en/stable/tutorials/notebooks/multimodal/totalVI.html) v1.1.2
- [sciPENN](https://github.com/jlakkis/sciPENN) v1.0.0
- [StabMap](https://github.com/MarioniLab/StabMap) v0.1.8
- [MultiVI](https://docs.scvi-tools.org/en/stable/tutorials/notebooks/multimodal/MultiVI_tutorial.html) v1.1.2
  
Cross Integration (Dimension Reduction, Batch Correction, Clustering, Classification):
- [totalVI](https://docs.scvi-tools.org/en/stable/tutorials/notebooks/multimodal/totalVI.html) v1.1.2
- [scMoMaT](https://github.com/PeterZZQ/scMoMaT) v0.2.2
- [UnitedNet](https://github.com/LiuLab-Bioelectronics-Harvard/UnitedNet) Github Version: 3689da8
- [sciPENN](https://github.com/jlakkis/sciPENN) v1.0.0
- [Concerto](https://github.com/melobio/Concerto-reproducibility) Github Version: ab1fc7f
- [scMDC](https://github.com/xianglin226/scMDC/tree/v1.0.0) Github Version:  43b0c3a
- [StabMap](https://github.com/MarioniLab/StabMap) v0.1.8
- [UINMF](http://htmlpreview.github.io/?https://github.com/welch-lab/liger/blob/master/vignettes/UINMF_vignette.html) v2.0.1
- [scMM](https://github.com/kodaim1115/scMM) Github Version: c5c8579
- [MOFA+](https://biofam.github.io/MOFA2/) v1.6.0
- [Multigrate](https://multigrate.readthedocs.io/en/latest/index.html) v0.0.2

Cross Integration (Spatial Registration):
- [PASTE](https://github.com/raphael-group/paste) (both pairwise and centre versions) v1.4.0
- [PASTE2](https://github.com/raphael-group/paste2) Gihub Version: b71ec88
- [SPIRAL](https://github.com/guott15/SPIRAL/tree/main) v1.0
- [GPSA](https://github.com/andrewcharlesjones/spatial-alignment) v0.8
  
Note that the installation time for tools may vary depending on the method used. For more detailed information, please refer to the original publication. For built-in classification, the classification scripts are provided in their corresponding method folders within the [tools_scripts] directory. For additional modules (such as kNN, SVM, random forest, and MLP), the scripts are provided in the [classification] directory.

## Evaluation Pipeline

All evaluation pipelines can be found within the [`evaluation_pipelines`](evaluation_pipelines/) folder. Small demo datasets are stored in [`data`](data/). See [`docs/data_contracts.md`](docs/data_contracts.md) for expected input and output formats.

For spatial registration data, users are required to download it from [link](https://drive.google.com/drive/folders/1pS3GzJrCipB13Am3PVjCfzPtXgTHm7wv?usp=share_link), and then place it under `data/spatial/`.

## Dataset

The processed datasets can be downloaded from [link](https://www.dropbox.com/scl/fo/6lismnrtgt4jm4ogvdmx2/AAcw4oqSTNjoOGoTvPVZPVU?rlkey=izc0rghdcjc0tyv8hgt3a6yws&st=037efszp&dl=0).

## Shiny

Explore method performance in depth with our interactive [Shiny](http://shiny.maths.usyd.edu.au/scMultiBench/), designed for dynamic visualization of benchmark results.

## Citation

If you use this repository, benchmark scripts, processed results, or datasets, please cite:

Liu, C., Ding, S., Kim, H.J., Long, S., Xiao, D., Ghazanfar, S. & Yang, P. Multitask benchmarking of single-cell multimodal omics integration methods. Nature Methods 22, 2449-2460 (2025). https://doi.org/10.1038/s41592-025-02856-3

## License

This project is covered under the Apache 2.0 License.

scMultiBench

Explore Similar Repositories

Was this content helpful?

Recent Feedback

Download README