A3D3a's MVP (adaMVP)

This is the official codebase for adaMVP: Probabilistic graph-based model uncovers druggable vulnerabilities in major solid cancers. paper

What is A3D3a's MVP?

A3D3a’s MVP (Adaptive AI-Augmented Drug Discovery and Development Molecular Vulnerability Picker) is a novel graph-based, cooperativity-led Markov chain model, developed and maintained by Ying Zhu, Stephanie Schmidt, et al. Bissan Al-lazikani lab at the University of Texas MD Anderson Cancer Center. The algorithm exploits cooperativity of weak signals within a cancer molecular network to enhance the signal of true molecular vulnerabilities.

Workflow of A3D3a's MVP

Installation

adaMVP works with Python >= 3.8. Please make sure you have the correct version of Python installed pre-installation.

We highly recommend using an isolated python environment using conda or virtualenv.

1. Create python>=3.8 environment

   • Using venv: python -m venv ada_mvp
   • Using conda: conda create -n ada_mvp python=3.8

2. Activate environment

   • Using venv: source ada_mvp/bin/activate
   • Using conda: conda activate ada_mvp

3. After setting the environment, you could install adaMVP via pip:

pip install adaMVP

4. (optional) Setup jupyter lab

• Install the ipython kernel: pip install -U ipykernel
• Install jupyter lab: pip install jupyterlab
• Introduce the virtual environment to jupyter: python -m ipykernel install --user --name 'ada_mvp'
• Open/Start Jupyter by typing jupyter lab and select the created kernel ada_mvp

Preparing INPUTS

Preparing your seed genes input file (mandatory)

This csv file must include two column names 'Gene' and 'Freq'. The 'Gene' column should have a list of genes with official gene symbols (HGNC symbols), and the 'Freq' column should be a list of numeric numbers between 0 and 1 representing the altered freq of a gene within the populations. An example file can be downloaded at example input file.

The recommended number of seed genes is between 50-250. For data from multiple modalities, e.g. mutation, CNV, RNA, the 'Freq' can be set as the maximum altered frequency across modalities. Please read the method part of the manuscript for detailed explanations.

Run examples

Find first neighbors and build MVP model for prioritizing molecular vulerabilities

from adaMVP import mvp_build_graph as mbg

mbg.find_fn_and_pgm(save_directory = output_path,
            to_remove = ['TTN','MUC16'],
            altered_freq_file = 'TCGA_BRCA_DNA_altered_freq.csv',
            fn_num = 550,
            thre = 0.05,
            Wm = 0.5,
            alpha = 0.1,
            n_perm = 1000)

Parameters (mandatory)

• altered_freq_file: input file with altered freq for each gene in a csv file.

Optional Parameters

• save_directory: directory path for saving output files
• to_remove: Genes to be filtered from the seed genes, default=[]
• threshold: Threshold of FDR for the permutation test for finding first neighbors of the seed genes, default=0.05
• fn_num: Maximum number of first neighbors to be brought into the network, sorted by the FDR and number of neighbors in the seeds, default=550, recommended range: 200-600
• n_perm: Number of iterations for the permutation test, default=10000
• Wm: weight parameter on the self-loop of nodes of the Markov chain model, default=0.5, recommended range: 0.4-0.7
• alpha: cooperativity factor describing transition between nodes, default=0.1, recommended range: 0-0.2. Note that a large alpha and a large Wm may cause the negative transition matrix problem and lead to negative final probability of the nodes. If that happens, please decrease the alpha and Wm to avoid the negative transition matrix problem.

Output files

Two output files, 'first_neighbors.csv' and 'markov_output.csv' will be saved in the assigned directory output_path. To understand the results, please check the results_documentation

Tutorial

The tutorial for running the adaMVP pipeline can be found at

build graph and community detection