Welcome!

The purpose of this Python package is to create 3D phylogenetic trees with two axes of variation given suitable data in the common AnnData format. The differentiating factor between this package and functions like scipy.cluster.hierarchy.dendrogram() and sklearn.cluster.AgglomerativeClustering is that the dendrogram produced tracks the splitting/merging patterns of groups of taxa and individual taxa. Specifically, as used in the BioRxiv paper Cerebellar nuclei evolved by repeatedly duplicating a conserved cell type set, the phylogenetic tree created by this package tracks the merging of different subnuclei of the cerebellar nuclei while also tracking the merging of individual cell types within those nuclei.

An example of such a dendrogram is

A flattened version can be found in Fig. S22C and Fig. S23H of the linked paper above.

This package is composed of three main parts:

• agglomerate
• data
• metrics

The agglomerate package exposes methods to perform the agglomeration of a single phylogenetic tree given suitable data and hyperparameters, and a method to perform batch agglomeration over a range of hyperparameters and select the best tree according to any of the following metrics:

• Balanced Minimum Evolution (preferred)
• Minimum Evolution
• Maximum Parsimony

The data package exposes a data_loader that the user can define to import their data accordingly (possibly from multiple folders or online repositories) and into an AnnData object. The data_types are used internally by the agglomeration algorithm.

Finally, the metrics currently only provides the Spearman correlation coefficient to measure the distance between two data points, however any distance metric in the same form as the example provided may be added and used in the agglomeration program.

This package requires Python version 3.7 or greater, and the requirements provided in Pipfile and Pipfile.lock. Using pip, installation is as easy as running:

pip install 3dtrees-nbingo

If you have any questions for how to use this code or for how it was used in Cerebellar nuclei evolved by repeatedly duplicating a conserved cell type set, then please feel free to email me at [email protected]. Examples usage of this package can be found in the cn_evolution repository, which is the analysis code used to produce the figures in the linked paper.