VIOLIN

VIOLIN (Validating Interactions Of Likely Importance to the Network)

VIOLIN (Validating Interactions Of Likely Importance to the Network) is a tool used to automatically classify and judge literature-extracted interactions curated from machine readers by comparing them to existing models. This comparison can help identify key interactions for model extension.

Contents

• Functionality
• I/O
• Online Tutorial
• Offline Installation
• Package Structure
• Reproducibility
• Citation
• Funding
• Support

Functionality

• Identification: identifying the importance of the interactions for a user-defined cellular model
• Error Detection: finding the contradictions in interactions list
• Scoring: showing the quality of the interactions for a model

I/O

Input/Output

• exmaple/input - folder containing model files and automatically retrieved interactions lists.
• example/output - example/output/<interactions_filename>_.csv, in BioRECIPE format, containing classified interactions (corroborations, contradictions, extensions, and flagged) and example/output/<interactions_filename>__score.csv (evidence score, match score, kind score, epistemic score and total score)
• use_VIOLIN.ipynb - Jupyter notebook used to run VIOLIN
• use_violin_script.py - python script used to run VIOLIN at the command line

I/O Annotations

Input Annotations	Query	Curation Method
RA1_reading.xlsx	Melanoma	REACH
RA2_reading.xlsx	MEK, ERK, AKT, GSK3, P70RSK, S6, CDK4, 4EBP1, YB1, SRC, CHK2, MTOR, PI3K	REACH explorer
RA2_0_1_reading.xlsx	RA2 without LEEs involving biological processes or chemicals	Manual
RA2_0_1_1_reading.xlsx	RA2_0_1 without LEEs which were redundant or irrelevant	Manual
RA3_reading.xlsx	MAPK/ERK pathway	REACH explorer
RA4_reading.xlsx	RPS6K1	REACH explorer

For each reading set R# in test_input

Output Annotations	Contents
R#_corroborations.csv	All LEEs judged as collaborations
R#_contradictions.csv	All LEEs judged as contradictions
R#_extensions.csv	All LEEs judged as extensions
R#_flagged.csv	All LEEs judged as flagged
R#_outputDF.csv	All LEEs, in order of descending Total Score

Online Tutorial

Run the demonstrated example; or alternatively upload user-customized input files (see I/O) to the input/ directory on File Browser Tab (upper left corner) of Binder.

This interactive jupyter notebook walks you though all of the code and functions to:

Offline Installation

1. Clone the VIOLIN repository to your computer.

   git clone https://github.com/pitt-miskov-zivanov-lab/VIOLIN.git
   

2. Navigate into the directory, install VIOLIN and its python dependencies.

   cd VIOLIN
   pip install -e .
   

3. Run the provided notebook (Check Jupyter notebook installation here).

   jupyter notebook examples/use_VIOLIN.ipynb
   

Package Structure

• setup.py: python file that help set up python dependencies installation
• src/violin/: directory that includes core python VIOLIN files
  • src/violin/formatting.py: functions of preprocessing strings in tabular input;
  • src/violin/in_out.py: functions of reading interactions list and model file and writing VIOLIN outputs;
  • src/violin/network.py: functions of creating model network and finding paths between nodes;
  • src/violin/scoring.py: implementation of decision tree for classification;
  • src/violin/visualize_violin.py: functions of visualizing classifying results;
• examples/: directory that includes tutorial notebook and example inputs and outputs
• environment.yml: environment file, required by Binder
• docs/: containing files supporting the repo's host on Read the Docs
• LICENSE.txt: MIT License
• README.md: this is me!

Reproducibility

The data and code for running experiments and creating figures in paper can be found in data.zip. The code is tested by example/test and example/test_VIOLIN.

Citation

@article{luo2024context,
  title={Context-driven interaction retrieval and classification for modeling, curation, and reuse},
  author={Luo, Haomiao and Hansen, Casey and Telmer, Cheryl A and Tang, Difei and Arazkhani, Niloofar and Zhou, Gaoxiang and Spirtes, Peter and Miskov-Zivanov, Natasa},
  journal={bioRxiv},
  pages={2024--07},
  year={2024},
  publisher={Cold Spring Harbor Laboratory}
}

Funding

This work was funded in part by DARPA Big Mechanism award, AIMCancer (W911NF-17-1-0135) and in part by the NSF EAGER award CCF-2324742.

Support

Feel free to reach out via email nmzivanov@pitt.edu for additional support if you run into any error.