Clustering method based on Forest Fire Dynamics
Project description
forest-fire-clustering
Clustering Method Inspired by Forest Fire Dynamics
Description
Forest Fire Clustering is an efficient and interpretable clustering method for discovering and validating cell types in single-cell sequencing analysis. Different than the existing methods, our clustering algorithm makes minimal prior assumptions about the data and provides point-wise posterior probabilities for internal validation. Additionally, it computes point-wise label entropies that can highlight novel transition cell types de novo along developmental pseudo-time manifolds. Lastly, our inductive algorithm is able to make robust inferences in an online-learning context.
Getting Started
Dependencies
- python >= 3.6
- numpy
- scipy
- scikit-learn
- numba
Installing
Estimated time: 2 mins
pip install forest-fire-clustering
or
pip install -u forest-fire-clustering
Executing program
"X" is a [sample x feature] matrix
To generate clustering:
import forest_fire_clustering.FFC as FFC
cluster_obj = FFC(X, sigma=0.1, k=300, num_permute=int(X.shape[0]/10))
cluster_obj.preprocess()
cluster_obj.fit(fire_temp=100)
To validate the results:
cluster_obj.validate()
cluster_obj.entropy()
cluster_obj.pval()
cluster_obj.entropy_list # list of entropies of the data point
cluster_obj.pval_list # list of posterior significance values
Authors
Zhanlin Chen, Jeremy Goldwasser, Philip Tuckman, Jing Zhang, Mark Gerstein
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Hashes for forest_fire_clustering-0.0.10.tar.gz
Algorithm | Hash digest | |
---|---|---|
SHA256 | 3c9af26f09887b108d86613bff49e681c4b56cf303274d13c2610ede4278cba8 |
|
MD5 | 865c9647e12c45d5c756aec54d2fc036 |
|
BLAKE2b-256 | c6dc9c7976b95025a8a3447829c89cc6bf43b225a6fd3d3e255153ee35d393a8 |
Hashes for forest_fire_clustering-0.0.10-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | d825c7000e401ccb863cffd06e4be8899b8d4389daf9f84c1cb54adc8a2ff7bf |
|
MD5 | d5485056e616fcd4cfa70855391e49d7 |
|
BLAKE2b-256 | 673b03ee7c72dd08cb5989c493504e1e7e621311169b0c96ffffa450a1215f69 |