RxGRID (Rapid proXimity Guidance for Repurposing Investigational Drugs) is a method for prioritizing candidates for drug repurposing based on network proximity to genetic hits from high-throughput screens. For additional documentation, please refer to our manuscript, available here.
RxGRID is built in Python 3 and requires NetworkX, NumPy, pandas, Matplotlib, and seaborn, in addition to the standard Python library. It was tested using Python 3.7, NetworkX 2.2, NumPy 1.21.6, pandas 1.2.3, Matplotlib 3.5.3, and seaborn 0.12.2. We provide an optional conda environment containing all necessary dependencies, which can be activated by running the following:
$ conda env create -f RxGRID_environment.yml
$ conda activate RxGRID
Clone the repository as follows:
$ git clone https://github.com/henrycousins/RxGRID
$ cd RxGRID
RxGRID can be run through the command line as follows:
python RxGRID.py --rnk_folder RNK_FOLDER --output_folder OUTPUT_FOLDER
where RNK_FOLDER is the name of the folder containing the collection of screens (as rnk files) to be analyzed and OUTPUT_FOLDER is the name of the existing folder to which the results should be saved. For instance, the following code performs RxGRID on the set of screens contained in /screens, saving the results to /results:
python RxGRID.py --rnk_folder screens --output_folder results
Additional parameters can be defined by the user.
usage: RxGRID.py [-h] [--gene_hit_threshold GENE_HIT_THRESHOLD]
[--drug_hit_threshold DRUG_HIT_THRESHOLD]
[--centrality_metric CENTRALITY_METRIC]
[--mutual_hit_freq MUTUAL_HIT_FREQ]
[--drug_exclusions_path DRUG_EXCLUSIONS_PATH]
[--rnk_folder RNK_FOLDER] [--output_folder OUTPUT_FOLDER]
[--save_plots SAVE_PLOTS]
optional arguments:
-h, --help show this help message and exit
--gene_hit_threshold GENE_HIT_THRESHOLD
Significance cutoff for genes (default=0.05)
--drug_hit_threshold DRUG_HIT_THRESHOLD
Significance cutoff for drugs (default=0.01)
--centrality_metric CENTRALITY_METRIC
Primary centrality metric for ranking drugs (degree
["ndc"] or betweenness ["btw"]; default="ndc"
--mutual_hit_freq MUTUAL_HIT_FREQ
Fraction of screens in which drug must be hit to meet
overall significance (inputs >=1 are interpreted as
counts, rather than fractions; default = 1/3)
--drug_exclusions_path DRUG_EXCLUSIONS_PATH
CSV file containing user-specified list of drugs to
exclude from downstream analysis
(default="data/drug_exclusions.csv")
--rnk_folder RNK_FOLDER
Name of folder containing the screens, as rnk files
(default="screens")
--output_folder OUTPUT_FOLDER
Name of folder to which results should be saved
(default="results")
--save_plots SAVE_PLOTS
Whether to save descriptive plots (default=False)
| File | Type | Extension | Columns | Comments |
|---|---|---|---|---|
RNK_FILE |
Input | .rnk | [gene ID, gene score] | Description |
<jobID>_summary |
Output | .csv | [screenwise hit count, hit status {per-screen}, percentile rank {mean, std, per-screen}, centrality score {mean, std, per-screen}] | Created automatically |
This software is available under an MIT License.
We anticipate releasing core updates on a 6-month basis for a minimum of 2 years and will respond to any issues regarding the codebase, which is open-source, as they arise. Supporting funding for RxGRID is provided by the National Institutes of Health (GM007365, GM102365, LM013337, and HG011316), by the Donald and Delia Baxter Foundation, by the National Science Foundation (1953686 and 1953415), and by the Knight-Hennessy Scholarships.