Skip to main content
This is a pre-production deployment of Warehouse. Changes made here affect the production instance of PyPI (
Help us improve Python packaging - Donate today!

Find nodes in hierarchical clustering that are statistically significant

Project Description

Author Haibao Tang (tanghaibao)
  Jingping Li (Jingping)
License BSD


Hierarchical clustering is an important tool in mining useful relationships among multivariate biological data. However, there is no obvious way to define a set of useful, non-overlapping groups from the identified hierarchy. Most efforts have focused on different cut-off values, evaluate the relative strengths of intra- versus inter- group variances and then heuristically determine a “good” cutoff. This study introduces a more dynamic approach that extracts clades that are significantly enriched or different from other clades. Incorporating phylogenetic information removes the false positives observed in a conventional analysis thus improves the prediction of trait association.

The algorithm takes two inputs, a tree model and some mapping of values for all the terminal branches. Briefly, the algorithm performs independent statistical tests on all the internal branches, and calculates the P-values for each node. At exploratory stage, the statistical tests are: 1) for quantitative values, test the difference of two groups separated by each node (student’s t-test); 2) for categorical values, test the association of a particular category for the descendants of each internal node (Fisher’s exact test).

The candidate nodes are determined using the following rule: the P-value for the candidate node v has to be the smallest among all root-to-leaf paths that pass v. In other words, the group rooted at node v should contain the largest level of association, thus avoiding redundant clades.

Server version

A server version of TREECUT software can be found here:


  • Python version >= 2.6
  • scipy for t-test and Fisher’s Exact Test
  • ete2 for parsing the tree structure
pip install scipy ete2


Take a look at examples in the data/ folder: treefile and listfile.

The treefile should be a Newick-formatted file (typically from the output of a phylogenetic reconstruction software, e.g. phylip or MEGA).

The listfile should contain the quantitative value for each taxon (separated by comma). Make sure that the taxon names match between treefile and listfile:

# continuous example

If the data type is discrete, separate the classes by semicolon. For example:

# discrete example

Note that # represents a comment line and will be ignored.

To run the software:

python data/tree.nwk data/continuous.csv tree.pdf

A summary of extracted modules will be written to stdout. Each row will contain a subclade that show either significantly high phenotypic value or low phenotypic value. Further a visualization is available as tree.pdf (supported image formats include svg, png, pdf, jpg, etc.). The modules are highlighted in green (low-value modules) and red (high-value modules) colors.

tree-value mapping


There are several immediate applications of TREECUT. Below just show case two examples, but there are more.

Extract taxonomic groups with high/low phenotype values

See an example in the data/ folder. This is the flowering time data for sorghum diversity panel. flowering.nwk is a phylogenetic tree for the sorghum accessions used in the study. flowering.assoc has the mapping to the accession to the trait values (in this case the number of days until flowering). To run:

python data/flowering.nwk data/flowering.assoc

If you stead want to treat the flowering data as discrete values, say “high” versus low. You can add a --discrete option:

python data/flowering.nwk data/flowering_discrete.assoc --discrete flowering_discrete.png

The significant different clades (like extreme trait values) will be written to the screen.

Extract co-expressed genes with functional enrichment

In this example, I used Eisen’s CLUSTER software (here) to process a series of arabidopsis microarray series AtGenExpress. After the CLUSTER is run. I found two files - microarray.cdt and microarray.gtr. The .gtr file contains a hierarchical tree structure, but I need to convert it to .nwk format in order for to process.

Take a look at microarray.assoc, this contains the mapping from arabidopsis genes to the GO terms, which are based on the information downloaded at Gene Ontology website. Note that a gene can have multiple GO terms associated with it. Here is the script that I used to create the microarray.assoc:

python scripts/

Once everything is set, just run as usual (make sure to turn on the --discrete option):

python scripts/ data/microarray.gtr data/microarray.cdt data/microarray.nwk
python data/microarray.nwk data/microarray.assoc --discrete

The clades that are significantly enriched in certain GO terms will be written to the screen.


Tang et al. TREECUT: algorithm for extracting significant modules from hierarchical clustering

Release History

This version
History Node


History Node


History Node


Download Files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Filename, Size & Hash SHA256 Hash Help File Type Python Version Upload Date
(12.6 kB) Copy SHA256 Hash SHA256
Source None Oct 15, 2017

Supported By

Elastic Elastic Search Pingdom Pingdom Monitoring Dyn Dyn DNS Sentry Sentry Error Logging CloudAMQP CloudAMQP RabbitMQ Kabu Creative Kabu Creative UX & Design Google Google Cloud Servers Fastly Fastly CDN StatusPage StatusPage Statuspage DigiCert DigiCert EV Certificate