drugbank-database-access

What it does

Parses the local DrugBank XML to provide drug information lookups by DrugBank ID, name, or CAS, including descriptions, indications, mechanisms, and groups. It can extract drug-drug interactions with descriptions, classify severity, and check specific drug pairs. It retrieves targets, enzymes, transporters, and carriers with UniProt accessions, as well as pathway associations. It fetches chemical properties (SMILES, MW, logP) and supports cross-database mappings to PubChem, ChEMBL, KEGG, and UniProt. It can compute molecular similarity using RDKit fingerprints between two drugs. It also builds a cross-reference table for multiple drugs. It emphasizes using the local XML for batch work and mentions an optional REST API with a 3,000 requests/month limit for live data.

How it works

• Loads the DrugBank XML with Python's ElementTree, using the namespace http://www.drugbank.ca for all XPath queries.
• Builds an index mapping DrugBank IDs and lowercase names to their drug entries; supports a CAS-number fallback search.
• get_drug_info(drug_element) extracts fields: drugbank_id, name, type, description, indication, mechanism_of_action, cas_number, and groups.
• search_by_name(pattern) returns drugs whose names contain the pattern (case-insensitive).
• get_interactions(drug_element) collects all interactions with drug names and descriptions; classify_severity maps description text to major/moderate/minor based on keywords.
• check_interaction(drug1_elem, drug2_elem) finds a specific interaction between two drugs and annotates its severity.
• get_targets(drug_element, target_type) extracts targets, enzymes, transporters, or carriers, including UniProt IDs and gene names, plus actions.
• get_pathways(drug_element) collects SMPDB pathway associations with SMPDB IDs and names.
• get_property(drug_element, kind_name, section) retrieves a single property value (e.g., SMILES, Molecular Weight) from calculated or experimental sections; get_all_properties(drug_element) aggregates all such properties.
• drug_similarity(drug1_elem, drug2_elem, radius, nbits) computes Tanimoto similarity using Morgan fingerprints from RDKit if SMILES are available.
• get_external_ids(drug_element) extracts cross-references like PubChem, ChEMBL, KEGG, and UniProt IDs.
• build_crossref_table(names) constructs a dataframe listing drug name, DrugBank ID, PubChem, ChEMBL, and KEGG identifiers for a set of names.

When to use it

• When looking up drug information by DrugBank ID, name, or CAS number.
• When checking drug-drug interactions and their severity classifications for polypharmacy safety.
• When extracting drug targets, enzymes, transporters, and carriers with UniProt accessions.
• When retrieving chemical properties (SMILES, InChI, MW) for cheminformatics tasks.
• When mapping DrugBank entries to PubChem, ChEMBL, UniProt, KEGG.
• When building drug similarity matrices from molecular fingerprints.
• For live bioactivity data or compound property lookups without downloading a database, use chembl-database-bioactivity or pubchem-compound-search respectively.

What it can touch

• Uses local XML file drugbank_all_full_database.xml
• Requires Python packages: lxml, rdkit (rdkit-pypi), pandas
• Optional: REST API access (not the primary data path) with a limit of 3,000 requests per month for developers
• External identifiers include PubChem, ChEMBL, KEGG, UniProt

Caveats

• License is NOASSERTION; XML namespace handling is mandatory for XPath queries to work correctly.
• Prerequisites require DrugBank account and XML download; performance depends on XML size and parsing approach (iterparse recommended for memory-constrained environments).
• Methods rely on textual descriptions to classify interaction severity, which may be approximate depending on input content.
• The description notes a primary local XML workflow and an optional REST API with rate limits; it does not guarantee live data accuracy beyond those constraints.