PROPKA is an empirical method for predicting the pKa values of ionizable amino acid residues in proteins based on their three-dimensional structure. Developed by Jan H. Jensen and colleagues at the University of Copenhagen, PROPKA calculates how the protein environment shifts the pKa of each ionizable group away from its intrinsic (model) value.
The pKa of an ionizable group determines at what pH that group gains or loses a proton. In solution, amino acids have characteristic pKa values, but within a folded protein, factors such as burial from solvent, hydrogen bonding, and electrostatic interactions with nearby charged residues can shift these values substantially. PROPKA quantifies these perturbations using structure-based empirical energy functions, enabling rapid prediction of protonation states at any given pH.
PROPKA 3 introduced consistent treatment of internal and surface residues through improved desolvation and dielectric response modeling. The method handles ASP, GLU, HIS, CYS, TYR, LYS, ARG residues and both N- and C-terminal groups. Benchmarks show root-mean-square deviations of 0.79 pKa units for Asp/Glu, 0.75 for Tyr, 0.65 for Lys, and 1.00 for His residues compared to experimental measurements.
How to use PROPKA online
ProteinIQ provides a web-based interface for running PROPKA 3.5.1 without command-line installation. Upload a protein structure file or fetch one from RCSB, and receive predicted pKa values for ionizable residues and supported ligand groups present in the structure.
Inputs
Input
Description
Protein Structures
The target protein structure. Upload a PDB file (.pdb or .ent) or enter a PDB ID to fetch from RCSB. Structures may include bound ligands if present in the PDB file. Supports batch processing of up to 10 structures.
Advanced settings
ProteinIQ supports a subset of safe native PROPKA options without changing default behavior:
Restrict to chains — Analyze only the specified comma-separated chain IDs
Titrate only residues — Limit titration to specific chain:resnum residue pairs
Show coupled residues — Include alternative pKa values for coupled sites
Keep input protons — Preserve hydrogens already present in the uploaded structure
Results
The output includes a spreadsheet of detected ionizable groups plus a Files tab with two download types:
A combined CSV export of the same columnar table shown in the spreadsheet
Raw PROPKA report files for each analyzed structure
Those raw reports contain the full native text output, including the folding profile and charge profile sections that PROPKA writes after the summary table.
Column
Description
Structure
The source structure identifier, typically the fetched PDB ID or uploaded filename stem.
Residue number for protein groups, or atom label for ligand groups.
Chain
Chain identifier from the PDB file.
pKa
Predicted pKa value for this residue in the protein environment.
Model pKa
Intrinsic pKa of the isolated amino acid in solution.
Shift
Difference between predicted and model pKa (ΔpKa=pKa−Model pKa).
Ligand atom type
native PROPKA ligand group/type annotation when the row comes from a ligand-derived titratable site. N/A means the row is a standard protein/terminal group with no ligand annotation.
Coupling note
native note describing discarded or coupled alternative pKa values when applicable. N/A means PROPKA did not report a coupling note for that row.
Why some fields show N/A
ProteinIQ now renders optional PROPKA fields as N/A instead of leaving them blank:
Ligand atom type = N/A — The row came from a standard protein residue or terminal group, not a ligand-derived titratable group
Coupling note = N/A — PROPKA did not attach a coupling/discard note to that row
N/A is intentional and means "not applicable here," not "missing due to a parser error."
What appears in Files
propka-results.csv — Combined export of the normalized results table for all analyzed structures
<structure>-propka-report.txt — Raw native PROPKA text report for each structure
If a field appears as N/A in the CSV, it has the same meaning as in the spreadsheet: the field was not applicable for that row.
Model pKa reference values
PROPKA uses the following intrinsic pKa values as baselines:
Residue
Model pKa
ASP (Aspartic acid)
3.80
GLU (Glutamic acid)
4.50
C-terminus
3.20
HIS (Histidine)
6.50
CYS (Cysteine)
9.00
TYR (Tyrosine)
10.00
LYS (Lysine)
10.50
ARG (Arginine)
12.50
N-terminus
8.00
Interpreting pKa shifts
The Shift column indicates how the protein environment alters the intrinsic pKa:
Positive shift — The residue is more difficult to ionize (stabilized neutral form). Common causes include burial from solvent, loss of favorable interactions in the charged state, or proximity to like-charged residues.
Negative shift — The residue ionizes more readily (stabilized charged form). Typically results from hydrogen bonds that stabilize the charged state or proximity to oppositely charged groups.
Large shifts (> 2 pKa units) — May indicate functionally important residues, catalytic sites, or unusual structural environments. Active site residues often display perturbed pKa values that enable catalysis at physiological pH.
Determining protonation states
To determine if a residue is protonated at a given pH:
Acidic residues (ASP, GLU, CYS, TYR, C-terminus) — Protonated (neutral) when pH < pKa; deprotonated (charged) when pH > pKa
Basic residues (HIS, LYS, ARG, N-terminus) — Protonated (charged) when pH < pKa; deprotonated (neutral) when pH > pKa
At physiological pH (~7.4), residues with predicted pKa values near 7.4 may exist in mixed protonation states.
How does PROPKA work?
PROPKA computes pKa values by calculating a perturbation to the model (intrinsic) pKa caused by the protein environment. The predicted pKa is expressed as:
pKa=pKamodel+ΔpKaDS+ΔpKa
where DS represents desolvation effects, HB represents hydrogen bonding interactions, and CC represents Coulombic charge-charge interactions.
Desolvation effects
When an ionizable group is buried within a protein, it loses favorable solvation interactions with water. Desolvation destabilizes the charged form relative to the neutral form, shifting carboxylate pKa values upward (making them harder to deprotonate) and amine pKa values downward (making them harder to protonate).
PROPKA calculates the desolvation penalty from two factors:
Solvent accessible surface area — Groups with less exposed surface experience greater desolvation penalties
Depth of burial — Distance from the protein surface, with deeply buried groups experiencing the largest effects
The desolvation contribution scales continuously between surface and internal residues, avoiding the discontinuous behavior of earlier PROPKA versions that treated these as discrete categories.
Hydrogen bonding
Hydrogen bonds can stabilize either the charged or neutral form of an ionizable group, depending on the geometry and donor/acceptor relationship:
Hydrogen bonds to charged carboxylates — Stabilize the charged form and shift pKa downward
Hydrogen bonds from protonated groups — Can stabilize the neutral form of acids
The hydrogen bonding contribution depends on both distance and angle, with optimal geometry producing the largest pKa shifts. PROPKA evaluates both side-chain and backbone hydrogen bond donors and acceptors.
Coulombic interactions
Electrostatic interactions between ionizable groups affect their pKa values:
Like charges — Repulsion destabilizes simultaneous ionization, shifting pKa values to reduce the probability of both groups being charged
For buried residue pairs, PROPKA applies full Coulombic interactions. For surface-exposed pairs, the high dielectric of water screens these interactions substantially. The transition between these regimes uses the same smooth interpolation applied to desolvation effects.
Coupled residues
When two ionizable groups interact strongly, their protonation states become coupled. Neither group can be assigned a single pKa value because the ionization of one depends on the protonation state of the other. PROPKA identifies these cases and marks them with an asterisk in detailed output.
Limitations
Static structure assumption — PROPKA analyzes a single conformation and does not account for conformational flexibility or dynamics that may affect pKa values
Empirical parameterization — Accuracy depends on the training set; unusual structural environments not represented in the parameterization may yield less reliable predictions
Coupled residues — Strongly interacting ionizable groups require special treatment; predicted pKa values for coupled pairs should be interpreted with caution
Ligand handling — PROPKA 3.1+ supports protein-ligand complexes, but ligand ionizable groups may be less accurately predicted than protein residues
Crystal packing artifacts — Structures from X-ray crystallography may contain crystal contacts that perturb pKa values in non-physiological ways
Arginine predictions — ARG predictions are less validated due to limited experimental data at high pH
Structures with missing atoms or poor resolution may produce unreliable results. Using PDB Fixer to repair structures before running PROPKA is recommended.
