Multitask benchmarking of single-cell multimodal omics integration methods,
with multibench: a Python API that runs 36 integration methods across
four categories (vertical, diagonal, mosaic, cross), scores them with scIB
metrics, and draws scIB-style bubble tables.
Documentation and tutorials: https://dsichang.github.io/scMultiBench/
Liu C, Ding S, Kim HJ, 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
Each method you run has its own paper; please cite it alongside the
benchmark. print(mtb.cite("Matilda", "MOFA2")) prints the benchmark's
reference and one line per method (multibench cite Matilda MOFA2 for BibTeX).
The benchmark and the method scripts live in PYangLab/scMultiBench.