MiTo (Mitochondrial single-cell lineage tracing Toolkit)

Mitochondrial single-cell multi-omics in Python.

MiTo infers clonal ancestries in single-cell data from natural mtDNA variation.

The framework provides the infrastructure to convert MAESTER, scmtATAC-seq and RedeeM pre-processing outputs into Allele Frequency Matrices (i.e., AnnData objects), filter informative mtDNA variants, genotype individual cells, and infer mitochondrial phylogenies and clones for downstream multi-omic analysis.

Versions. This README documents MiTo 1.0.0, the stable release of the original MiTo API, the one used for the analyses of the MiTo paper. MiTo 2.x has a redesigned API and one pipeline for MAESTER, ReDeeM and mtscATAC: see the current documentation. To stay on this release, use pip install "scmito<2". See CHANGELOG.md.

Installation

pip install scmito

Requires Python 3.11+.

import mito as mt
print(mt.__version__)

Quick start

import scanpy as sc
import mito as mt

# Allele Frequency Matrix (cell x site AnnData with AD / DP layers)
afm = sc.read('afm_unfiltered.h5ad')

# Cell and variant filters, distances in mtDNA mutation space
afm = mt.pp.filter_cells(afm, cell_filter='filter2')
afm = mt.pp.filter_afm(afm, filtering='MiTo')

# Phylogeny and clonal inference
tree = mt.tl.build_tree(afm, precomputed=True, solver='UPMGA')
annotator = mt.tl.MiToTreeAnnotator(tree)
annotator.clonal_inference()

# Visualization
mt.pl.plot_tree(tree, features=['MiTo clone'])

See getting-started tutorial for the full vignette.

API

MiTo follows the scverse layout, composing with scanpy and anndata:

Supported platforms: scRNA-seq (MAESTER), scmtATAC-seq and 10x MultiOme (RedeeM).

Full reference: MiTo docs.

Reproducing the paper

The paper analyses run with MiTo 1.0.0, using the conda recipe in MiTo_benchmark_repro (envs/environment.yml, which installs scmito==1.0.0 and the additional libraries the scripts need).

The Nature Communications release is 0.1.5. 1.0.0 differs from it in one way that matters for these analyses: mt.pp.kNN_graph(from_distances=True) returns the k nearest neighbours, where 0.1.5 returned every other cell. kNN-based quantities (kNN purity, entropy and kBET, UMAP embeddings, Leiden clustering on distances) therefore differ from the published ones. In the comparison we ran against the deposited tuning table, distance- and tree-based metrics (AUPRC, ARI, NMI) were identical. See CHANGELOG.md.

Development

git clone https://github.com/andrecossa5/MiTo.git
cd MiTo
pip install -e ".[test]"
pytest

Citation

If MiTo is useful in your work, please cite:

Cossa, A. Dalmasso A. et al. MiTo: mitochondrial lineage tracing and single-cell multi-omics. Nat Comm (2026). https://doi.org/10.1038/s41467-026-71607-5

MiTo builds on the Cassiopeia package for phylogeny reconstruction and data infrastructure. Please, cite it as well.

Releases

See CHANGELOG.md.

License

MIT — see LICENSE.