Updated for CovPDB 2.0

Help & FAQ

Everything you need to browse, search, and download data from CovPDB — a freely accessible database of experimentally determined and computationally predicted covalent protein–ligand interactions, curated from high-resolution structures in the Protein Data Bank.

Getting started

The fastest way into the database is straight from the homepage.

The CovPDB homepage shows six representative icons — Ligands, Proteins, Complexes, Warheads, Mechanisms, and Residues. Clicking any of them opens the complete, sortable list for that data category. The same lists are always available from the navigation bar under Browse.

CovPDB homepage with the six quick-access icons for ligands, proteins, complexes, warheads, mechanisms, and residues — The six quick-access icons on the CovPDB homepage.

At a glance, CovPDB 2.0 currently contains:

Covalent complexes	6,186
Pre-reactive ligands	3,616
Target proteins	5,844
Reactive warheads	119
Covalent mechanisms	27
Targetable residues	15
Covalently bound complexes (modeled by AF3)	6,783
Non-covalently bound complexes near cysteines	7,109

Browse CovPDB

Each data category has its own browsable, sortable table. All entries are cross-linked, so you can move freely between ligands, proteins, complexes, warheads, mechanisms, and residues.

Ligands

Home›Browse›Ligands

The ligand list shows every pre-reactive ligand in the database together with its 2D structure and the number of covalent complexes it appears in.

Sortable list of ligands with structures and complex counts — The ligand list with structures and complex counts.

Sort the list alphabetically by Ligand Name, or by the number of Complexes.
Click a Ligand ID to open the ligand card with full details for that compound.

Proteins

Home›Browse›Proteins

Sort the protein list by Protein Name, UniProt ID, Organism, or the number of Complexes.
Click any protein name to open its protein card.
Each protein card lists up to three tables: Covalently Bound Ligands, Covalently Bound Ligands (modeled by AF3), and Non-covalently Bound Ligands near to Cysteines.

New in CovPDB 2.0: in addition to experimentally determined complexes, protein cards now include AlphaFold 3 (AF3) predicted complex models for ligandable cysteine residues (integrated from chemoproteomics data in CysDB) and non-covalent binders located near reactive cysteines — useful starting points for designing new covalent ligands.

Upper part of a protein card with general protein information — Protein card — general information and structure overview.

Lower part of a protein card with tables of bound ligands — Protein card — tables of covalently and non-covalently bound ligands.

Complexes

Home›Browse›Complexes

Each row summarizes a complex: protein name, UniProt ID, ligand name, and HET code.
Click View to open the complex card with the covalent bonding details and an interactive 3D visualization of the protein–ligand structure.

Complex card with covalent bonding details and 3D structure viewer — A complex card, including the interactive 3D structure viewer.

Warheads

Home›Browse›Warheads

The warhead table lists all 119 reactive warheads that form covalent bonds with targetable residues, each defined by a SMARTS pattern and shown with its 2D structure.

Warhead table with names, SMARTS patterns, and structures — Warhead names, SMARTS patterns, and structures.

Warhead table with counts of related ligands, proteins, mechanisms, residues, and complexes — Counts of ligands, proteins, mechanisms, residues, and complexes per warhead.

Each warhead entry shows its name, SMARTS pattern, and structure.
The table also lists how many ligands, proteins, mechanisms, residues, and complexes are associated with each warhead.
Every count is a link — click it to drill down. For example, click the Residues count of a warhead and then a Ligands count to see which ligands target that residue.

Covalent mechanisms

Home›Browse›Cov. Mechanisms

A table displays all 27 covalent bonding mechanism types annotated in CovPDB 2.0.

Table of covalent mechanism types — The covalent mechanism table.

Detailed view of a covalent mechanism with its reaction scheme — Detailed view of a single mechanism.

Click a Mechanism Name to open its detailed view, including the reaction scheme.
From there, explore all ligands, proteins, warheads, residues, and complexes associated with that mechanism.

Targetable residues

Home›Browse›Targetable Residues

A table displays all 15 amino acid residues that are covalently targeted in the database.

Table of targetable residues — The targetable residue table.

Related data linked from a targetable residue.

For each residue, the table shows the associated ligands, proteins, warheads, mechanisms, and complexes.
Click any count to open the corresponding list for the residue of interest.

↑ Back to top

Search CovPDB

CovPDB 2.0 offers three search modes, all available from the Search menu: keyword search, chemical similarity / substructure search, and protein sequence search.

Keyword search

Home›Search›Keyword

Use the keyword search whenever you already know a name or identifier:

Ligand — search by ligand name, synonym, HET code, or CovPDB ligand ID. Matches open the corresponding ligand card.
Protein — search by protein name or synonym, or by UniProt ID / accession. Matches open the corresponding protein card.
PDB ID — enter a four-character PDB ID to jump directly to the complex card of that PDB entry.

Keyword search form for ligand names, protein names, and PDB IDs — The keyword search interface.

Similarity / structure search

Home›Search›Similarity/Structure

When a compound’s name is unknown or ambiguous but its structure is known, draw the structure directly in the integrated ChemDoodle editor (or paste a SMILES string). An exact match is not required — structurally similar compounds are retrieved as well.

Similarity search: lower the Tanimoto coefficient threshold to retrieve more (less similar) compounds, or raise it for stricter matches.
Substructure search: retrieve every database compound containing your query as a substructure — for example, draw a tricyclic scaffold to find all tricyclic ligands in CovPDB.

Ligand similarity search with the ChemDoodle structure editor — Structure-based searching with the ChemDoodle editor.

Similarity searches are performed with RDKit using Morgan fingerprints — binary bit arrays that encode the structural features of a molecule. Similarity between two fingerprints is quantified by the Tanimoto coefficient:

tc_A,B = N_A,B / (N_A + N_B − N_A,B)

N_A — number of bits set “on” in the fingerprint of molecule A
N_B — number of bits set “on” in the fingerprint of molecule B
N_A,B — number of bits set “on” in the fingerprints of both molecules (A and B)

A Tanimoto coefficient of 0.85 (85%) is widely accepted as a reasonable first guess that two molecules may exhibit comparable biological activities.

Protein sequence search

Home›Search›Protein Sequence

Sequence Similarity Searching of CovPDB uses BLAST+ tool to perform fast local alignment searches for a query sequence. Type in a query sequence without header.

The search result page displays an overview of all the retrieved hits with their corresponding scoring. Below the table, a more detailed alignment result is displayed.

Protein sequence search form — The protein sequence search interface.

↑ Back to top

Download

Home›Download

All CovPDB content is freely available for bulk download as flat files:

All ligand structures — SDF format
All protein sequences — FASTA format
All protein–ligand complexes — PDB format (ZIP archive)
All reactive warheads — TXT format (SMARTS definitions)
AlphaFold 3 (AF3) predicted complex models for ligandable cysteine residues (new in 2.0)

Download page listing the available data set files — The download page with all available data sets.

Linked resources

CovPDB cross-references and cites data from the following public resources:

Protein Data Bank (PDB)
Protein Data Bank in Europe (PDBe)
Universal Protein Resource (UniProt)
ChEMBL
PubChem
DrugBank
NCBI Gene
BindingDB
PDBbind
Binding MOAD
CysDB — chemoproteomics-derived ligandable human cysteine residues (new in 2.0)

FAQs

What is new in CovPDB 2.0?

CovPDB 2.0 nearly triples the number of curated covalent protein–ligand complexes compared to the first release. Beyond the expansion, the update adds:

Integration of chemoproteomics-derived ligandable human cysteine residues from CysDB;
Complex models for ligandable cysteine residues predicted with AlphaFold 3 (AF3);
Predicted covalent binding sites identified through large-scale structural pocket analysis;
Improved interactive 3D visualization for proteins and ligands;
A protein sequence search mode;
An expanded download section, including the AF3 predicted complex models.

What is the CovPDB ID?

The CovPDB ID is a unique identifier assigned to each ligand in CovPDB. It can be used in the keyword search to retrieve the ligand card of that compound directly.

What are “modeled” complexes and “non-covalently bound ligands near cysteines”?

In addition to experimentally determined covalent complexes from the PDB, CovPDB 2.0 provides covalently bound complexes (modeled by AF3) — protein–ligand complex structures predicted with AlphaFold 3 (AF3) for ligandable cysteine residues — and non-covalently bound ligands near to cysteines, i.e. non-covalent binders whose binding sites lie close to a reactive cysteine. Both data sets support the systematic identification of novel covalent druggable targets and the design of new covalent ligands.

How often is the data updated?

CovPDB is updated periodically to include new entries as newly deposited 3D structures become available in the PDB.

How can CovPDB support my research?

CovPDB was the first freely accessible web database solely dedicated to high-resolution 3D structures of biologically relevant covalent protein–ligand complexes mined from the Protein Data Bank. Version 2.0 describes 6,186 covalent complexes, 3,616 pre-reactive ligands, 5,844 target proteins, 119 reactive warheads, 27 covalent mechanisms, and 15 targetable residues. Typical applications include:

Mechanistic studies: the curated warheads and high-resolution structures are well suited for QM or QM/MM studies of enzyme reactions and warhead reactivity.
Structure-based covalent ligand design: thousands of co-crystal structures provide templates for designing covalent ligands against a protein of interest. The complexes in the Download section can be used directly as input for structure-based design workflows and AI-based structure prediction methods such as AlphaFold 3.
Target discovery: the CysDB-derived ligandable cysteines, AF3-predicted complex models, and predicted covalent binding sites in 2.0 support the systematic identification of novel covalent druggable targets.
Sequence analyses: the FASTA download enables studies of conserved sequence motifs and homology modeling.

How were the SASA and pKₐ values of the targetable residues calculated?

SASA and pK_a values were calculated with the FreeSASA and PROPKA programs, respectively.

FreeSASA is an open-source library for calculating solvent-accessible surface areas (SASA) of protein, RNA, and DNA molecules. SASA values of the targetable residues were computed directly from the PDB files using default parameters.

PROPKA (version 3) predicts the local pK_a values of ionizable groups of the targetable amino acid residues, accounting for desolvation effects and the dielectric response of the protein.

How should ABPP-derived evidence be labeled?

Ligandability labels were assigned using the 4S-4R consensus criterion, where nS-mR denotes the minimum numbers of supporting sources (S) and records (R), respectively. The source and record counts used for label assignment were taken from the LigCysABPP database [Link]. The LigCysABPP authors identified 4S-4R as the criterion offering the best balance between label confidence and data coverage [DOI:10.1101/2025.09.07.670677v3].

Cysteines meeting the 4S-4R criterion were labeled high confidence: liganded cysteines were classified as ligandable, and unliganded cysteines as unligandable. Cysteines that did not meet the criterion were labeled low confidence—ambiguous. Cysteines not represented in LigCysABPP were labeled low confidence—unavailable.

LigCysABPP is a chemoproteomic database built from peptide-level measurements reported in 15 cysteine-directed ABPP studies published between 2016 and 2025 [DOI:10.1101/2025.09.07.670677v3].

Cite us

If CovPDB is useful for your work, please cite:

Mingjie Gao, Aurélien F. A. Moumbock, Ammar Qaseem, Qianqing Xu, Stefan Günther.
CovPDB: a high-resolution coverage of the covalent protein–ligand interactome.
Nucleic Acids Research, 2022, 50(D1), D445–D450. doi:10.1093/nar/gkab868

Questions or feedback? Reach us via the Contact page.