About SurfPotDB

What this is

SurfPotDB is a public, citable database of protein surface electrostatic potentials, computed by solving the linearized Poisson–Boltzmann equation with NextGenPB on structures from the RCSB Protein Data Bank. It is freely accessible (no login, no registration), fully functional, and queryable by PDB code.

What makes SurfPotDB different?

SurfPotDB is designed to make protein electrostatics reproducible and reusable at database scale. Instead of requiring users to recompute surface potentials with their own protonation choices, dielectric constants, ionic conditions and meshing parameters, SurfPotDB provides a consistently generated reference set.

Each entry is computed from an explicitly protonated structure, using the same documented physical parameters across the database. The downloadable files include both the PQR structure and the enriched VTP surface mesh, allowing users not only to inspect the surface potential but also to reuse the data for downstream calculations without rerunning the Poisson–Boltzmann solver.

What SurfPotDB gives you:

• One documented parameter set for every structure. Solute dielectric 2, solvent 80, ionic strength 0.145 M, 298.15 K, so values are directly comparable across the whole database.
• The exact protonation behind every value. Protonation is assigned at pH 7.0 with MCCE4 and published as a .pqr file carrying per-atom charges and radii, with no hidden assumptions.
• The field anywhere inside the protein, from the files alone. The surface meshes (.vtp) are enriched with polarization charges, so you can reconstruct the electrostatic potential and field at any interior point with no solver re-run. A small standalone tool is provided.
• Persistent, citable releases. All data are deposited on Zenodo under a persistent concept DOI, including a canonical CSV index and a SQLite snapshot for offline, dataset-wide analysis.

Quick start

1. Enter a PDB code in the search box.
2. If the structure is a computed representative, SurfPotDB shows its electrostatic quantities and download links.
3. If the structure belongs to a covered 90%-identity cluster, SurfPotDB redirects you to the computed representative for that cluster.
4. Download the VTP file to visualize the surface electrostatic potential in ParaView/PyVista, or for analysis in Python.
5. Download the PQR file to inspect the protonation state, per-atom charges and radii, or to reuse it in other electrostatics workflows.
6. Optional: use the provided tool to reconstruct the electrostatic potential and field on individual atoms and at any point inside the structure, directly from the published files.

Current scope

The present release focuses on the ligand- and ion-free fraction of the non-redundant PDB, so that every entry can be computed with a single, well-defined charge and radius protocol. Starting from 126,139 structures grouped into 32,861 non-redundant clusters (90% sequence identity), it provides electrostatics for the 8,571 protein-only representatives, which answer for roughly 19,000 PDB codes.

The remaining ~24,000 clusters contain bound ligands or ions. Protein–ligand and protein–ion complexes are planned for future releases, once robust ligand and ion parametrization is incorporated.

How broad it is

The computed representatives are non-redundant by construction and together span a broad swath of the known protein universe, across families, folds, functions and the tree of life:

Full per-class breakdowns, figures and the selection, protonation and calculation protocols are on the Methods page.

Access & reuse

• Open and immediate. Browse and query by PDB code with no account; member codes of a covered cluster resolve automatically to the computed representative.
• Two files per structure. The surface potential mesh (.vtp, for ParaView/PyVista) and the protonated structure (.pqr, for PyMOL/VMD), plus a dataset-wide CSV index and SQLite snapshot.
• Citable. Persistent Zenodo DOIs.

What's next

The database is actively maintained and growing:

• Protein–ligand and protein–ion complexes, extending the electrostatics beyond protein-only entries.
• Neural-network–refined surface potentials, trained on analytical benchmarks for higher accuracy.
• More representatives, toward all 32,861 clusters, with regular DOI releases that add structures without breaking existing download links.

← Back to the database  ·  Methods & data details →