clep 0.0.3

A Hybrid Data and Knowledge Driven Framework for Generating Patient Representations

CLEP: A Hybrid Data- and Knowledge- Driven Framework for Generating Patient Representations

Table of Contents

• General Info
• Installation
• Documentation
• Input Data
• Usage
• Issues
• Acknowledgements
• Citation
• Disclaimer

General Info

CLEP is a framework that contains novel methods for generating patient representations from any patient level data and its corresponding prior knowledge encoded in a knowledge graph. The framework is depicted in the graphic below

Installation

The code can be installed from PyPI with:

$ pip install clep

The most recent code can be installed from the source on GitHub with:

$ pip install git+https://github.com/hybrid-kg/clep.git

For developers, the repository can be cloned from GitHub and installed in editable mode with:

$ git clone https://github.com/hybrid-kg/clep.git
$ cd clep
$ pip install -e .

Documentation

Read the official docs for more information.

Input Data Formats

Data

Symbol	Sample_1	Sample_2	Sample_3
HGNC_ID_1	0.354	2.568	1.564
HGNC_ID_2	1.255	1.232	0.26452
HGNC_ID_3	3.256	1.5	1.5462

Note: The data must be in a tab separated file format.

Design

FileName	Target
Sample_1	Abnormal
Sample_2	Abnormal
Sample_3	Control

Note: The data must be in a tab separated file format.

Knowledge Graph

The graph format CLEP can handle is a modified version of the Edge List Format. Which looks as follows:

Source	Relation	Target
HGNC_ID_1	association	HGNC_ID_2
HGNC_ID_2	decreases	HGNC_ID_3
HGNC_ID_3	increases	HGNC_ID_1

Note: The data must be in a tab separated file format & if your knowledge graph does not have relations between the source and the target, just populate the relation column with "No Relation".

Usage

Note: These are very basic commands for clep, and the detailed options for each command can be found in the documentation

1. Radical Searching The following command finds the extreme samples with extreme feature values based on the control population.

$ clep sample-scoring radical-search --data <DATA_FILE> --design <DESIGN_FILE> --control Control --threshold 2.5 --control_based --ret_summary --out <OUTPUT_DIR>

2. Graph Generation The following command generates the patient-gene network based on the method chosen (Interaction_network).

$ clep embedding generate-network --data <SCORED_DATA_FILE> --method interaction_network --ret_summary --out <OUTPUT_DIR>

3. Knowledge Graph Embedding

The following command generates the embedding of the network passed to it.

$ clep embedding kge --data <NETWORK_FILE> --design <DESIGN_FILE> --model_config <MODEL_CONFIG.json> --train_size 0.8 --validation_size 0.1 --out <OUTPUT_DIR>

4. Classification

The following command carries out classification on the given data file for a chosen model (Elastic Net) using a chosen optimizer (Grid Search).

$ clep classify --data <EMBEDDING_FILE> --model elastic_net --optimizer grid_search --out <OUTPUT_DIR>

Issues

If you have difficulties using CLEP, please open an issue at our GitHub repository.

Acknowledgements

Citation

If you have found CLEP useful in your work, please consider citing:

CLEP: A Hybrid Data- and Knowledge- Driven Framework for Generating Patient Representations.
Bharadhwaj, V. S., Ali, M., Birkenbihl, C., Mubeen, S., Lehmann, J., Hofmann-Apitius, M., Hoyt, C. T., & Domingo-Fernandez, D. (2020).
Bioinformatics, btab340.

Graphics

The CLEP logo and framework graphic was designed by Carina Steinborn.

Disclaimer

CLEP is a scientific software that has been developed in an academic capacity, and thus comes with no warranty or guarantee of maintenance, support, or back-up of data.