TMAP2

Tree-based visualization for high-dimensional data. Organizes similar items into interactive tree structures. Ideal for chemical space, protein embeddings, single-cell data, or any high-dimensional dataset.

Why Trees?

Most dimensionality reduction tools (UMAP, t-SNE, PCA) produce point clouds. TMAP produces a tree, a connected structure where every point is linked to its neighbors through branches. This makes the layout itself explorable because you can follow branches, trace paths between any two points, and discover how regions connect.

For example, in a TMAP of pet breed images, following the branch from terriers toward cats reveals that the bridge between the two groups runs through chihuahuas and sphynx cats (the bald ones) which is both hilarious and logical since both are small, have short hair, big eyes... The tree doesn't just cluster similar things it also shows you how dissimilar things are connected.

Because the layout is a tree, you get operations that point clouds can't support:

path = model.path(idx_a, idx_b)        # nodes along the tree path
d = model.distance(idx_a, idx_b)        # sum of edge weights along the path
n = model.hops(idx_a, idx_b)            # number of tree edges between two points
pseudotime = model.distances_from(idx)  # tree distance from one point to all others

The same path tracing and tree-distance colouring are available inside the interactive HTML: pin two points and the inspector lists the path, or colour the whole map by tree distance from a selected point.

Installation

pip install tmap2

Wheels are published for Linux x86_64, macOS arm64 (Apple Silicon), and Windows x86_64 on Python 3.11 to 3.13. Other platforms build the OGDF layout extension from source and need CMake and a C++17 compiler.

Optional extras:

pip install rdkit # chemistry helpers (fingerprints_from_smiles, molecular_properties)
pip install jupyter-scatter # notebook interactive widgets
pip install biopython # protein helpers (ProtParam properties, PDB parsing)

Note: The import name is tmap, not tmap2.

Quick Start

Binary Data (e.g. Chemical Fingerprints)

from tmap.utils import fingerprints_from_smiles
from tmap import TMAP

smiles = [...]  # your SMILES list
# Binary fingerprints (Jaccard distance). `valid` flags the SMILES RDKit could parse,
# so you can drop the same rows from any labels or properties you attach later.
fps, valid = fingerprints_from_smiles(smiles, fp_type="morgan", radius=2, n_bits=2048, return_valid=True)
model = TMAP(metric="jaccard", n_neighbors=20).fit(fps)
viz = model.to_tmapviz()
viz.write_html("map.html")  # interactive HTML, open in a browser
# viz.show()                # or render inline in a Jupyter notebook

Continuous Vectors (e.g. Protein Embeddings)

import numpy as np
from tmap import TMAP

# embeddings (use cosine / euclidean distances)
X = np.random.random((1000, 128)).astype(np.float32)
model = TMAP(metric="cosine", n_neighbors=20).fit(X)
viz = model.to_tmapviz()
viz.show()                    # inline in a Jupyter notebook
# viz.write_html("tmap.html") # or save as interactive HTML

Key Features

• Tree structure: follow branches, trace paths, count hops, compute pseudotime
• Always one tree: a too-low n_neighbors can fragment the kNN graph; TMAP bridges the pieces so path and distance stay defined (connect_components=True, inspect with n_components_)
• Deterministic: the layout is always seeded and deterministic. For cosine/euclidean, pass reproducible=True to also make the HNSW index build bit-identical across runs (slower)
• Multiple metrics: jaccard, cosine, euclidean, precomputed, or bring your own kNN graph
• Incremental: add_points() and transform() for adding new data into an existing TMAP
• Model persistence: save() / load()
• Three viz backends: interactive HTML, jupyter-scatter, matplotlib

Visualization (add colors, labels...)

Notebook widgets: color switching, categorical filtering, and lasso selection with pandas-backed metadata:

Add Colors & Labels

Adding colors is quite simple. Just pass the name of the layout (e.g. Molecular Weight, Age, Protein Lenght ...), a list of values for each node and matplotlib color. If the data is categorical (e.g. Age or Heavy Atom Count) pass categorical=True so that categorical colors like tab10 become available. To add labels (i.e. data that is not needed for coloring the nodes) just pass a name for the labels and the list of values.

model = TMAP(metric="jaccard").fit(X)
viz = model.to_tmapviz() 
viz.add_color_layout("Molecular Weight", mw.tolist(), categorical=False) 
viz.add_color_layout("Scaffold", scaffolds, categorical=True, color="tab10")
viz.add_label("SMILES", smiles_list)
viz.show(width=1000, height=620, controls=True) # to see in jupyter notebook
# viz.write_html("mytmap.html") # to save and see as HTML in the browser

Here SMILES are added as a plain label, so no 2D structure is drawn. To render structures in tooltips and cards, use viz.add_smiles(smiles_list) instead. For image datasets use viz.add_images(paths_or_urls).

Filters, cards and structures

viz.add_filter("Ring Count", n_rings, categorical=True)   # filter-panel column without a colour map
viz.configure_column("UniProt ID", link_template="https://www.uniprot.org/uniprotkb/{value}")
viz.configure_card(title_column="Name", fields=["Molecular Weight", "Scaffold"])
viz.add_3d_structures(alphafold_urls, source="url", fmt="pdb")   # or add_3d_structure_files(local_paths)

add_filter puts a column in the filter panel without computing colours for it, which is cheaper than add_color_layout when you only want to filter. Colour layouts are always filterable. configure_column and configure_card control links, formatting and what the pinned card shows. add_3d_structures and add_3d_structure_files attach PDB or mmCIF structures that render in the card.

Interactive HTML

viz.write_html("name.html") writes a self-contained page with lasso selection, light/dark theme, filter and search panels, pinned metadata cards, and a binary mode for large datasets. Selecting a point opens the inspector:

• Neighbors: the point's tree neighbours with similarity scores, property differences and structures or images. Hovering a neighbour highlights the connecting edge.
• Path: pin a second point to trace the tree path between them, listing each node with its step number and running tree distance.
• Colour by tree distance: colour the whole map by tree distance from the selected point, using the normal colour menu.

For publication figures use matplotlib: model.plot_static(color_by=labels).

Domain Utilities

Built-in helpers for common scientific workflows:

from tmap.utils.chemistry import fingerprints_from_smiles, molecular_properties
from tmap.utils.proteins import fetch_uniprot, sequence_properties
from tmap.utils.singlecell import from_anndata

Examples

Runnable scripts for chemistry, images, proteins and text live in examples/. The shortest one is:

python examples/chemistry/molecules_tmap.py --nrows 3000

Notebooks

Lower-Level Pipeline

For direct control over indexing, hashing, and layout, see the legacy pipeline notebook. The main building blocks:

from tmap.index import USearchIndex           # dense / binary kNN
from tmap import MinHash, LSHForest           # Jaccard on sets / strings
from tmap.layout import LayoutConfig, layout_from_lsh_forest

Your Data
   ├─→ Binary matrix ─────────→ USearch        (Jaccard / cosine / euclidean)
   └─→ Sets / strings ───────→ MinHash → LSHForest
                ↓
             k-NN Graph → MST → OGDF Tree Layout → Interactive Visualization

Development

git clone https://github.com/afloresep/tmap2.git
cd tmap2
pip install ".[dev]"
pytest -v

License

MIT License; see LICENSE for details.

Citation

If you use TMAP2 in your research, please cite:

From Proteins and Molecules to Cats and Dogs: Visualization as Scalable Minimum Spanning Trees
Alejandro Flores Sepúlveda, Maarten Boneschansker, Daniel Probst, Jean-Louis Reymond
ChemRxiv, 2026.
https://doi.org/10.26434/chemrxiv.15008307/v1

@article{floressepulveda2026tmap,
  title   = {From Proteins and Molecules to Cats and Dogs: Visualization as Scalable Minimum Spanning Trees},
  author  = {Flores Sepúlveda, Alejandro and Boneschansker, Maarten and Probst, Daniel and Reymond, Jean-Louis},
  year    = {2026},
  doi     = {10.26434/chemrxiv.15008307/v1},
  url     = {https://doi.org/10.26434/chemrxiv.15008307/v1},
  note    = {ChemRxiv preprint}
}