Skip to content

Latest commit

 

History

History
119 lines (96 loc) · 6.15 KB

README.md

File metadata and controls

119 lines (96 loc) · 6.15 KB
Raw

SARS-CoV-2-network-analysis

Analysis of SARS-CoV-2 molecular networks

Getting Started

  • Required Python packages: networkx, numpy, scipy, pandas, sklearn, pyyaml, rpy2, wget, tqdm, graphspace_python
  • Required R packages: PPROC
  • Recommended R packages: clusterProfiler, org.Hs.eg.db

We recommend using Anaconda for Python, especially to access the needed R packages. To setup your environment, use the following commands:

conda create -n sarscov2-net python=3.7 r=3.6
conda activate sarscov2-net
pip install -r requirements.txt

To install the R packages:

R -e "install.packages('https://cran.r-project.org/src/contrib/PRROC_1.3.1.tar.gz', type = 'source')"
conda install -c bioconda bioconductor-clusterprofiler

Download Datasets

The SARS-CoV-2 - Human PPI "Krogan" dataset is available in the datasets/protein-networks folder.

To download additional datasets and map various namespaces to UniProt IDs, use the following command

python src/masterscript.py --config config-files/master-config.yaml --download-only

The YAML config file contains the list of datasets to download and is self-documented. The following types of datasets are supported:

  • Networks
  • Gene sets
  • Drug targets

To download additional datasets, copy one of the existing dataset sections in the config file and modify the fields accordingly. If your dataset is not yet supported, add to an existing issue (#5), or make an issue and we'll try to add it as soon as we can.

Run the FastSinkSource Pipeline

This master script will generate a config file specific to the FastSinkSource pipeline, which will then be used to generate predictions, and/or run cross validation.

Note that the FastSinkSource code was added as a git-subrepo, so to make changes to that code, please commit them to that repository directly, and then pull them with git subrepo pull src/FastSinkSource/, or follow the suggestions in the git-subrepo documentation.

Generate predictions

The script will automatically generate predictions from each of the given methods in the config file. The default number of predictions stored is 10. To write more, either set the num_pred_to_write or fator_pred_to_write flags under fastsinksource_pipeline_settings -> eval_settings, or, after the config file is generated, call the run_eval_algs.py script directly and add either the --num-pred-to-write or --factor-pred-to-write options (see python src/FastSinkSource/run_eval_algs.py --help)

Example 1, with the appropriate flags set in the master config file:

python src/masterscript.py --config config-files/master-config.yaml

Example 2:

python src/FastSinkSource/run_eval_algs.py  \
  --config fss_inputs/config_files/stringv11/400-nf5-nr100.yaml \
  --num-pred-to-write -1

Test for enrichment of top predictions with various gene sets (#6)

After the predictions have been generated, you can test for the enrichment of various genesets among the top k predictions.

The following command will test for enrichment of the top 332 predictions of each algorithm, and on each dataset/network in the config file:

python src/Enrichment/fss_enrichment.py \
    --config fss_inputs/config_files/stringv11/400-nf5-nr100.yaml \
    --k-to-test 332 --file-per-alg

TODO Currently only tests for enrichment of GO terms (BP, MF, CC).

To test for enrichment of any given list of genes (e.g., Krogan nodes), use the following type of command:

python src/Enrichment/enrichment.py \
  --prot-list-file fss_inputs/pos-neg/2020-03-sarscov2-human-ppi/2020-03-24-sarscov2-human-ppi.txt \
  --prot-universe-file fss_inputs/networks/stringv11/400/sparse-nets/c400-node-ids.txt \
  --out-pref outputs/enrichment/krogan \
  --add-prot-list-to-prot-universe

Post sub-network of the top predictions and their scores to GraphSpace

https://graphspace.org/ allows you to visualize and interact with the predictions made by the methods. In addition, you can share graphs with a group to easily allow others to do the same.

Here's an example of how to post the top 5 predictions made by GM+ when using STRING, and the shortest paths from those to the virus proteins:

python src/graphspace/sars_cov2_post_to_gs.py \
  --config fss_inputs/config_files/stringv11/400-nf5-nr100.yaml \
  --sarscov2-human-ppis datasets/protein-networks/2020-03-biorxiv-krogan-sars-cov-2-human-ppi.tsv \
  --user <email> --pass <password> \
  --k-to-test 5 --name-postfix=-test \
  --alg genemaniaplus \
  --parent-nodes

Cross Validation

Similar to the previous section, the options to run cross validation can either be set in the config file under fastsinksource_pipeline_settings -> eval_settings, or passed directly to run_eval_algs.py. The relevant options are below. See python src/FastSinkSource/run_eval_algs.py --help for more details.

  • cross_validation_folds
    • Number of folds to use for cross validation. Specifying this parameter will also run CV
  • sample_neg_examples_factor
    • ratio of negatives to positives to randomly sample from the nodes of the given network
  • num_reps
    • Number of times to repeat CV and the sampling of negative examples
  • cv_seed
    • The seed to use when generating the CV splits.

Example:

python src/FastSinkSource/run_eval_algs.py \
  --config fss_inputs/config_files/stringv11/400-nf5-nr100.yaml \
  --cross-validation-folds 5

After CV has finished, to visualize the results, use the plot.py script (TODO add to masterscript.py). For example:

python FastSinkSource/plot.py --config fss_inputs/config_files/stringv11/400-nf5-nr100.yaml --box --measure fmax

I used the jupyter notebook src/jupyter-notebooks/plot_net_collections.ipynb to visualize the CV results (e.g., Fmax) on the large collections of TissueNet v2 (TODO make a script for it).