Multiscale Interactome Data -- Revised

Original GitHub Repository: https://github.com/snap-stanford/multiscale-interactome

Forked GitHub Repository: https://github.com/callahantiff/multiscale-interactome/tree/development

This repository stores a revised version of the original data that was used in the publication titled: Identification of disease treatment mechanisms through the multiscale interactome. As described in the original GitHub's Readme, the original data can be downloaded directly from: http://snap.stanford.edu/multiscale-interactome/data/data.tar.gz. 

Description of Original Data

Drug-Protein (n=8,568):

• Source(s):
  • DrugBank (v5.1.1; 2018; drugbank_approved_target_uniprot_links.csv)
  • Drug Repurposing Hub (September 2018)
• Processing: map Uniprot to Entrez gene using HUGO (October 2018) and drug ids to DrugBank ids
• Filtering: Filter proteins to only keep those that appear in the Protein-Protein edge set.

Disease-Protein (n=25,212):

• Source(s): DisGeNet (March 2018)
• Filtering: only keep only expert curated gene-disease associations. (1) exclude disease-gene relationships that are inferred, based on orthology, animal models, or literature mining; (2) remove therapeutic disease-gene associations; and (3) remove disease-gene relationships that do not appear in the Protein-Protein edge set.

Protein-Protein (n=387,626):

• Source(s):
  • BioGRID (v3.5.178; November 2019; BIOGRID-ORGANISM-Homo_sapiens-3.5.178.tab)
  • Database of Interacting Proteins (February 2017; Hsapi20170205.txt). Include all experimental methods
  • Human Reference Protein Interactome Mapping Project. Four networks derived from high-throughput yeast two hybrid assays.
  • Menche 2015 (PMID:25700523). Compiles different types of physical protein-protein interactions.
• Processing: Map protein ids to Entrez gene ids using HUGO (sources 1-2 only)
• Filtering: only human proteins with physical interactions and direct experimental evidence (no genetic/indirect)

Protein-Biological Process (n=34,777): Source(s): Gene Ontology (human; February 2018)

• Processing: use master ids provided by GOATOOLS (v0.8.4)
• Filtering: only allow: EXP, IDA, IMP, IGI, HTP, HDA, HMP, HGI. Exclude any protein-biological functions inferred from: physical interactions, gene expression patterns, phylogenetically inferred annotations or computational analyses, automatic annotations (i.e., based on author statements, curator inference, electronic annotation), and those with no biological data

Biological Process-Biological Process (n=22,545):

• Source(s): Gene Ontology (human; February 2018) + Gene Ontology Plus (human version; July 2020)
• Filtering: Allow following relationship types: regulates, positively regulates, negatively regulates, part of, is a. Only consider BPs associated with at least one drug target or disease protein (directly or through children)

⚠️ Updates to Original Implementation ⚠️

Modifications to Original Data and Code

• Ensured every entry had a valid identifier and label
• Reconciled duplicate gene entries (i.e., gene identifiers that had been merged)
• Changed genes are listed in: resources/data/updated_gene_identifiers.xlsx