Description

This folder contains three sub folders corresponding to three different datasets. Each of these sub folders contain separate script for UniPath, AUCell, GSVA and PAGODA along with the data necessary for running these four methods. The scripts have been used for evaluating performance of UniPath with other three methods in terms of low dimensional visualization and clustering 

Requirements

R version > 3.6

In order to run these scripts user requires following packages:

1. UniPath
2. AUCell
3. GSVA
4. scde
5. Rtsne
6. Dbscan
7. aricode

Usage

Input

For MCA dataset and GGSE97930 UniPath takes UMI counts as input on the other hand fpkm data have been used for Tabula Muris dataset. For all other methods UMI counts and count data have been provided as an input. Along with the expression data, gene sets have been used as an input to transform expression data into pathway scores. 

Pathway scores are then subjected to Principal component analysis (PCA). Then, first 50 principal components are used as an input to Rtsne for low dimensional visualization.

This is followed by dbscan based clustering of tsne coordinates using different eps values which in the script have been labelled as n. User can change eps values as per their requirement. We have used eps values of 0.5,1,1.5, 2, 3 and 5. Then based on the dbscan clusters at different eps values, we compute ARI, NMI from aricode package in R and clustering purity for assessing clustering performance from all the four methods.

Output

tSNE plots and clustering performance measured in terms of ARI, NMI and clustering purity