PARADYS

In this repository we present our tool PARADYS for PAtient-specific RAnking of genes driving DYSregulations in cancer. Below you can find information on installation, usage and exemplary calls.

Installation

To use PARADYS you need to have Python installed on your system. Install the necessary conda environment by running

conda env create -f environment.yml

and then activate the newly created environment paradys_tool.

Usage

You can use PARADYS from the command line with the following arguments:

python paradys.py --patients <PATIENT_IDS> --networks <NETWORK_FILE> --mutations <MUTATION_FILE> --outdir <OUTPUT_DIR> --kappa <KAPPA> [--scores] [--all]

The input arguments are given by:

PATIENT_IDS: List of space-separated patient IDs of patients that are supposed to be analyzed. The patients IDs must match with those used in your input files.
NETWORK_FILE: Path to the input dysregulation networks file.
MUTATION_FILE: Path to the input mutation matrix file.
OUTPUT_DIR: Path to output directory for storing results.
KAPPA: Value of parameter kappa (int). Sets size of considered shortest-path neighborhood for putative driver gene identification. Larger values correspond to potentially more detected drivers, but also increase runtime.
--scores: Set this flag if you want to compute impact scores for drivers. (optional, default is False)
--all: Set this flag if you wish to analyze every patient in the given cohort (optional, default is False). If set, --patients flag can be omitted.

Input

PARADYS expects two comma-separated .csv input files:

The input NETWORK_FILE must contain the columns 'patient', 'tf', and 'gene'. It has to contain all directed dysregulation edges (a,b) from your personalized input networks, with the source node a stored in column 'tf' and target node b stored in 'gene'. The column 'patient' stores the ID of the patient that this edge belongs to.
The input MUTATION_FILE must contain the two columns 'patient' and 'gene'. It has to contain the sets of mutated genes for each patient, with the mutated gene in the column 'gene' and the corresponding patient ID in the column 'patient'.

In the directory preprocessing/ we present an exemplary script for generating such input data from TCGA mutation data and SSN networks.

Example Call

We added two toy input data files under example/networks_mini.csv and examples/mutations_mini.csv. For analyzing patients P2 and P3 you can then run PARADYS on this input by calling:

python paradys.py --patients P2 P3 --networks examples/networks_mini.csv --mutations examples/mutations_mini.csv --outdir results/ --kappa 2

For additionally computing per-patient driver impact scores you can simply run

python paradys.py --patients P2 P3 --networks examples/networks_mini.csv --mutations examples/mutations_mini.csv --outdir results/ --kappa 2 --scores

If you want to use PARADYS on "real-life" input data, follow the download instructions in preprocessing/README.md. Be aware that this input cohort consists of ~470 patients and therefore takes significantly longer to compute.

Output

In the output directory, for each analyzed patient we create a file named <PATIENT_ID>_drivers.csv storing detected significant driver genes for each analyzed patient. In the column 'driver' we store the detected significant driver gene, in combination with the associated dysregulation edge in the column 'dysregulation' and the respective p-value of the chi-squared test in the column 'pvalue'.

If additionally the --scores flag was set, we create a per-patient score file named <PATIENT_ID>_scores.csv storing drivers in the column 'driver' and the corresponding computed impact score in 'score'.