mTopic - Multimodal topic modeling for single-cell data

Contents

mTopic - Multimodal topic modeling for single-cell data#

mTopic is an open-source Python library with computational tools for modeling multimodal topics in single-cell data.

mTopic logo (light) mTopic logo (dark)

mTopic is a scalable Python package for topic modeling on multimodal single-cell datasets. It supports spatial and non-spatial data, enabling a joint analysis across multiple molecular layers, such as RNA, ATAC, and protein expression.

By capturing shared patterns across modalities, mTopic helps reveal latent biological structures, regulatory programs, and spatial organization within complex tissues.

Resources#

Features#

  • Supports simultaneous inference of multimodal molecular programs from any set of single-cell modalities

  • Handles spatial and non-spatial single-cell data

  • Implements scalable variational inference for efficient training

  • Allows per-modality preprocessing, signature detection, and z-score enrichment

  • Built on top of the MuData (https://muon.readthedocs.io/) data structure (.h5mu format)

  • Visualization tools for topic distributions, dominant topic maps, and feature signatures

Contents