xDTD Analysis

Overview

The xDTD Analysis repository is dedicated to the investigation and analysis of variances in KGML-xDTD models, with a focus on random forest (RF) prediction probabilty performance. This project is a part of the 6 week rotation project at David Koslicki Lab.

Rotation Project Objectives

1. Investigate Schema Variances Between the Models:

   • Explore and analyze differences between xDTD models.
   • Examine each model's schema characteristics and trends.

2. Investigate RF Prediction Probability Performance:

   • Examine RF prediction probability results for 100 drug-disease pairs.
   • Assess prediction performance variance for each model.

Repository Contents

Each scripts contain data location and instructions necessary to perform each analysis. To conduct the analysis use the scripts in the following order below:

1. xDTD_Schema_Analysis_Scripts contains scripts for schema analsysis for Node Mapping, Path Result, and Drug-Disease Pairs.

2. xDTD_Drug-DiseasePairs_Extract contains script to extract drug-disease pairs. The directory also contains 154 drug-disease pairs used for the rotation project in CSV file format.

3. xDTD_RandomForest_Instruction contains script to run the random forest mode to obtain the drug-disease prediction probabiltiy.

4. xDTD_RandomForest_Prediction_Analysis is the final step in analysis. It contains script to extract prediction probability scores of hundred drug-disease pairs and visualization, scatter plot, of the result.

Jupyter Notebook Script

The core of the analysis is encapsulated in the Jupyter notebook script. This script contains the necessary code and instructions to perform the KGML-xDTD analysis, addressing the rotation project objectives.

Figures

The folder holds the visual output results of the analysis. These results are presented in both PDF, PNG, and JPEG formats.

Pre-Setting

1. Install Conda.
2. Obtain .yml files for the conda environment from here.
3. Install conda environments using the following commands:

# Install Mamba
conda install -c conda-forge mamba

# Configure conda environments
mamba env create -f envs/graphsage_p2.7env.yml
mamba env create -f envs/xDTD_training_pipeline_env.yml

# Activiate the 'xDTD_training_pipeline' conda environment
conda activate xDTD_training_pipeline

Start of Analysis

To begin your analysis or reproduce the results, follow these steps:

1. Clone the repository to your local machine:

   git clone https://github.com/your-username/xDTD-Analysis.git
   

2. Navigate the project directory:

   cd xDTD-Analysis

3. Import, or download, and decompress the data files. The location of the data for each analysis is provided in each Jupyter Notebook script.

4. Open and run the Jupyter notebook scripts:

   jupyter notebook <name_of_the_script>.ipynb

This will initiate the analysis and generate the results.

Note:

We recommend you have the following specifications to successfuly run the pipeline and the scripts.

• Operating system(s): Linux (Ubuntu)

• Programming language: Shell Script (Bash) and Jupyter Notebook Script with Python 3.8.12

• Other requirements: Mamba version 1.5.0 and Anaconda version 23.7.4