BindingDB (https://www.bindingdb.org/) is the primary public database of measured drug-protein binding affinities. It contains roughly 3.2 million binding data records for ~1.4 million compounds tested against ~11,500 protein targets (homepage figures, 2026-09), curated from scientific literature and patent literature. BindingDB stores quantitative binding measurements (Ki, Kd, IC50, EC50) essential for drug discovery, pharmacology, and computational chemistry research.
\nKey resources:
\nresponse=application/json)BindingDB_All_<YYYYMM>_tsv.zip, ~560 MB zipped, refreshed monthly)Use BindingDB when:
\n| Scenario | \nUse Instead | \n
|---|---|
| Curated bioactivity mining at scale, assay metadata, drug mechanisms | \nalterlab-chembl | \n
| Compound identifiers/properties by name or CID, PubChem BioAssay | \nalterlab-pubchem | \n
| Purchasable analogs or docking-ready 3D libraries | \nalterlab-zinc-db | \n
| Docking a ligand into a receptor structure | \nalterlab-diffdock | \n
Base URL: https://bindingdb.org/rest
import requests\n\nBASE_URL = \"https://bindingdb.org/rest\"\n\ndef bindingdb_query(method, params):\n \"\"\"Query the BindingDB REST API.\"\"\"\n url = f\"{BASE_URL}/{method}\"\n response = requests.get(url, params=params, headers={\"Accept\": \"application/json\"})\n response.raise_for_status()\n return response.json()\nThe getLigandsByUniprot endpoint takes a single uniprot parameter formatted as\n<UniProt accession>;<cutoff in nM>, e.g. P00519;10000.
def get_ligands_for_target(uniprot_id, cutoff=10000):\n \"\"\"\n Get all ligands with measured affinity for a UniProt target.\n\n Args:\n uniprot_id: UniProt accession (e.g., \"P00519\" for ABL1)\n cutoff: Maximum affinity value to return (in nM)\n \"\"\"\n params = {\n \"uniprot\": f\"{uniprot_id};{cutoff}\",\n \"response\": \"application/json\",\n }\n return bindingdb_query(\"getLigandsByUniprot\", params)\n\n# Example: Get all compounds binding ABL1 (imatinib target) at <=100 nM\nligands = get_ligands_for_target(\"P00519\", cutoff=100)\n\n# Response shape (verified 2026-09): one top-level key, spelled\n# \"getLindsByUniprotResponse\" (sic — getTargetByCompound uses the same key),\n# holding \"bdb.hit\" (count, as a string) and \"bdb.affinities\": a list of\n# {\"bdb.monomerid\", \"bdb.smile\", \"bdb.affinity_type\", \"bdb.affinity\"}.\n# Affinities are strings, may carry leading spaces or \">\"/\"<\" qualifiers.\nresp = next(iter(ligands.values()))\nrows = resp.get(\"bdb.affinities\", [])\ndef search_by_smiles(smiles, cutoff=0.85):\n \"\"\"\n Search BindingDB by SMILES string (structural-similarity search).\n\n Args:\n smiles: SMILES string of the compound\n cutoff: Tanimoto similarity threshold (0.0-1.0; e.g. 0.85)\n \"\"\"\n params = {\n \"smiles\": smiles,\n \"cutoff\": cutoff,\n \"response\": \"application/json\",\n }\n return bindingdb_query(\"getTargetByCompound\", params)\n\n# Example: structural-similarity search for imatinib's binding targets\nresult = search_by_smiles(\"Cc1ccc(NC(=O)c2ccc(CN3CCN(C)CC3)cc2)cc1Nc1nccc(-c2cccnc2)n1\")\nTo query by a PubChem CID, first convert the CID to a SMILES string via PubChem\nPUG-REST, then pass it to search_by_smiles. There is no by-name or by-CID\nBindingDB REST endpoint.
For comprehensive analyses, download BindingDB data directly:
\nimport pandas as pd\n\ndef load_bindingdb(filepath=\"BindingDB_All.tsv\"):\n \"\"\"\n Load BindingDB TSV file (the unzipped BindingDB_All_<YYYYMM>.tsv).\n Download the dated BindingDB_All_<YYYYMM>_tsv.zip from:\n https://www.bindingdb.org/rwd/bind/chemsearch/marvin/Download.jsp\n \"\"\"\n # Key columns\n usecols = [\n \"BindingDB Reactant_set_id\",\n \"Ligand SMILES\",\n \"Ligand InChI\",\n \"Ligand InChI Key\",\n \"BindingDB Target Chain Sequence\",\n \"PDB ID(s) for Ligand-Target Complex\",\n \"UniProt (SwissProt) Entry Name of Target Chain\",\n \"UniProt (SwissProt) Primary ID of Target Chain\",\n \"UniProt (TrEMBL) Primary ID of Target Chain\",\n \"Ki (nM)\",\n \"IC50 (nM)\",\n \"Kd (nM)\",\n \"EC50 (nM)\",\n \"kon (M-1-s-1)\",\n \"koff (s-1)\",\n \"Target Name\",\n \"Target Source Organism According to Curator or DataSource\",\n \"Number of Protein Chains in Target (>1 implies a multichain complex)\",\n \"PubChem CID\",\n \"PubChem SID\",\n \"ChEMBL ID of Ligand\",\n \"DrugBank ID of Ligand\",\n ]\n\n # Exact TSV headers drift between monthly releases (some contain double\n # spaces), so intersect with the actual header rather than hard-failing.\n header = pd.read_csv(filepath, sep=\"\\t\", nrows=0).columns\n keep = [c for c in usecols if c in header]\n df = pd.read_csv(filepath, sep=\"\\t\", usecols=keep,\n low_memory=False, on_bad_lines='skip')\n\n # Convert affinity columns to numeric\n for col in [\"Ki (nM)\", \"IC50 (nM)\", \"Kd (nM)\", \"EC50 (nM)\"]:\n if col in df.columns:\n df[col] = pd.to_numeric(df[col], errors='coerce')\n\n return df\n\ndef query_target_affinity(df, uniprot_id, affinity_types=None, max_nm=10000):\n \"\"\"Query loaded BindingDB for a specific target.\"\"\"\n if affinity_types is None:\n affinity_types = [\"Ki (nM)\", \"IC50 (nM)\", \"Kd (nM)\"]\n\n # Filter by UniProt ID\n mask = df[\"UniProt (SwissProt) Primary ID of Target Chain\"] == uniprot_id\n target_df = df[mask].copy()\n\n # Filter by affinity cutoff\n has_affinity = pd.Series(False, index=target_df.index)\n for col in affinity_types:\n if col in target_df.columns:\n has_affinity |= target_df[col] <= max_nm\n\n result = target_df[has_affinity][[\"Ligand SMILES\"] + affinity_types +\n [\"PubChem CID\", \"ChEMBL ID of Ligand\"]].dropna(how='all')\n return result.sort_values(affinity_types[0])\nimport pandas as pd\n\ndef sar_analysis(df, target_uniprot, affinity_col=\"IC50 (nM)\"):\n \"\"\"\n Structure-activity relationship analysis for a target.\n Retrieves all compounds with affinity data and ranks by potency.\n \"\"\"\n target_data = query_target_affinity(df, target_uniprot, [affinity_col])\n\n if target_data.empty:\n return target_data\n\n # Add pIC50 (negative log of IC50 in molar)\n if affinity_col in target_data.columns:\n target_data = target_data[target_data[affinity_col].notna()].copy()\n target_data[\"pAffinity\"] = -((target_data[affinity_col] * 1e-9).apply(\n lambda x: __import__('math').log10(x)\n ))\n target_data = target_data.sort_values(\"pAffinity\", ascending=False)\n\n return target_data\n\n# Most potent compounds against EGFR (P00533)\n# sar = sar_analysis(df, \"P00533\", \"IC50 (nM)\")\n# print(sar.head(20))\ndef polypharmacology_profile(df, ligand_smiles, affinity_cutoff_nM=1000):\n \"\"\"\n Find all targets a compound binds to, by exact SMILES match.\n For tolerance to tautomers/salts/charge states, match on the InChIKey\n skeleton (first 14 chars of \"Ligand InChI Key\") instead of raw SMILES.\n \"\"\"\n # Search by ligand SMILES (exact string match)\n mask = df[\"Ligand SMILES\"] == ligand_smiles\n\n ligand_data = df[mask].copy()\n\n # Filter by affinity\n aff_cols = [\"Ki (nM)\", \"IC50 (nM)\", \"Kd (nM)\"]\n has_aff = pd.Series(False, index=ligand_data.index)\n for col in aff_cols:\n if col in ligand_data.columns:\n has_aff |= ligand_data[col] <= affinity_cutoff_nM\n\n result = ligand_data[has_aff][\n [\"Target Name\", \"UniProt (SwissProt) Primary ID of Target Chain\"] + aff_cols\n ].dropna(how='all')\n\n # Rank by the tightest measured constant per row (NaNs ignored)\n result = result.assign(best_nM=result[aff_cols].min(axis=1))\n return result.sort_values(\"best_nM\")\nimport pandas as pd\n\ndef find_best_inhibitors(uniprot_id, affinity_type=\"IC50 (nM)\", top_n=20):\n \"\"\"Find the most potent inhibitors for a target in BindingDB.\"\"\"\n df = load_bindingdb(\"BindingDB_All.tsv\") # Load once and reuse\n result = query_target_affinity(df, uniprot_id, [affinity_type])\n\n if result.empty:\n print(f\"No data found for {uniprot_id}\")\n return result\n\n result = result.sort_values(affinity_type).head(top_n)\n print(f\"Top {top_n} inhibitors for {uniprot_id} by {affinity_type}:\")\n for _, row in result.iterrows():\n print(f\" {row['PubChem CID']}: {row[affinity_type]:.1f} nM | SMILES: {row['Ligand SMILES'][:40]}...\")\n return result\ndef prepare_ml_dataset(df, uniprot_ids, affinity_col=\"IC50 (nM)\",\n max_affinity_nM=100000, min_count=50):\n \"\"\"Prepare BindingDB data for ML model training.\"\"\"\n records = []\n for uid in uniprot_ids:\n target_df = query_target_affinity(df, uid, [affinity_col], max_affinity_nM)\n if len(target_df) >= min_count:\n target_df = target_df.copy()\n target_df[\"target\"] = uid\n records.append(target_df)\n\n if not records:\n return pd.DataFrame()\n\n combined = pd.concat(records)\n # Add pAffinity (normalized)\n combined[\"pAffinity\"] = -((combined[affinity_col] * 1e-9).apply(\n lambda x: __import__('math').log10(max(x, 1e-12))\n ))\n return combined[[\"Ligand SMILES\", \"target\", \"pAffinity\", affinity_col]].dropna()\n| Field | \nDescription | \n
|---|---|
Ligand SMILES | \n2D structure of the compound | \n
Ligand InChI Key | \nUnique chemical identifier | \n
Ki (nM) | \nInhibition constant (equilibrium, functional) | \n
Kd (nM) | \nDissociation constant (thermodynamic, binding) | \n
IC50 (nM) | \nHalf-maximal inhibitory concentration | \n
EC50 (nM) | \nHalf-maximal effective concentration | \n
kon (M-1-s-1) | \nAssociation rate constant | \n
koff (s-1) | \nDissociation rate constant | \n
UniProt (SwissProt) Primary ID | \nTarget UniProt accession | \n
Target Name | \nProtein name | \n
PDB ID(s) for Ligand-Target Complex | \nCrystal structures | \n
PubChem CID | \nPubChem compound ID | \n
ChEMBL ID of Ligand | \nChEMBL compound ID | \n
| Affinity | \nClassification | \nDrug-likeness | \n
|---|---|---|
| < 1 nM | \nSub-nanomolar | \nVery potent (picomolar range) | \n
| 1–10 nM | \nNanomolar | \nPotent, typical for approved drugs | \n
| 10–100 nM | \nModerate | \nCommon lead compounds | \n
| 100–1000 nM | \nWeak | \nFragment/starting point | \n
| > 1000 nM | \nVery weak | \nGenerally below drug-relevance threshold | \n
>, <, or >= prefixes (e.g. >10000) for censored measurements. pd.to_numeric(errors='coerce') silently turns these into NaN — strip the prefix first (e.g. df[col].astype(str).str.lstrip(\"<>= \")) and decide explicitly whether to keep, drop, or treat censored values as inequalities before modelingTarget Source Organism to ensure human protein datascripts/query_bindingdb.py — runnable helper for the BindingDB REST API (no key):
python scripts/query_bindingdb.py uniprot P00519 --cutoff 10000\npython scripts/query_bindingdb.py pdb 1Q0L,3ANM --cutoff 100 --identity 92\npython scripts/query_bindingdb.py compound \"<SMILES>\" --cutoff 0.85\n