Pocket scores, druggability classification, residue lists, and PyMOL visualization scripts from P2Rank.

P2Rank returns a ranked list of predicted pockets, each with a probability score and a druggability class. The top-scoring druggable pocket is the site to target next — feed it to DiffDock as the docking target, or use its druggability to decide whether the protein is worth pursuing at all.

Output artifacts

Each completed P2Rank job returns a ranked list of predicted binding pockets with scores, druggability classification, contributing residues, and optional visualization data.

Primary output fields

Pocket table

Field	Type	Description
pocket_rank	integer	Rank by probability score (1 = most likely binding site)
pocket_score	float (0-1)	Probability of being a true binding site
druggability	string	Druggability classification (druggable, intermediate, difficult, undruggable)
residue_count	integer	Number of amino acid residues lining the pocket
residues	array	List of residue identifiers (e.g., ASP-189, HIS-57)
volume	float \| null	Always null: P2Rank 2.4.2 does not report pocket volume
surface_area	float \| null	Always null: P2Rank 2.4.2 does not report pocket surface area
center_x/y/z	float	Geometric center of the pocket in angstroms
hydrophobicity	float	Fraction of hydrophobic residues in the pocket
polarity	float	Fraction of polar residues in the pocket

Pocket score interpretation

Score Range	Confidence	Interpretation
0.8-1.0	Very high	Strong binding site prediction; likely a known or novel functional pocket
0.5-0.8	High	Probable binding site; suitable for docking and further analysis
0.3-0.5	Moderate	Possible binding site; may be a secondary or allosteric site
<0.3	Low	Unlikely binding site; may be a surface concavity without functional relevance

Druggability classification

The druggability class is a platform heuristic, not a P2Rank output. It combines the P2Rank pocket score (most of the weight) with the pocket's hydrophobic and aromatic residue fractions. Because P2Rank reports no pocket volume, volume is not part of the score, and each result carries a druggability_caveat saying so:

Classification	Description	Typical Properties
Druggable	High confidence that the pocket can bind drug-like small molecules	High P2Rank score, mixed hydrophobic/polar lining
Intermediate	Potentially druggable with optimized compounds	Moderate P2Rank score, variable lining
Difficult	Challenging to drug with standard small molecules; may require specialized approaches	Low P2Rank score or highly polar lining
Undruggable	Very unlikely to bind drug-like molecules effectively	Very low P2Rank score

Druggability considerations

Druggability is assessed for conventional drug-like small molecules (MW 200-500, Lipinski-compliant)
Fragments (MW < 200) can bind to pockets classified as "difficult" or even "undruggable"
PPI inhibitors and macrocycles may target shallow pockets that P2Rank classifies as difficult
Covalent binders can target pockets near reactive residues regardless of druggability score

Residue analysis

Pocket-lining residues

Each pocket includes a list of amino acid residues that form the pocket surface:

Residue format: Three-letter code with chain and number (e.g., ASP-A-189)
Key residues: Catalytic residues, hydrogen bond donors/acceptors, and hydrophobic contacts
Residue types: Distribution of hydrophobic, polar, charged, and aromatic residues

Interpreting residue composition

Pattern	Significance
Mix of hydrophobic and polar residues	Characteristic of druggable binding pockets
Predominantly hydrophobic	Likely binds hydrophobic ligands; typical of lipid-binding sites
Many charged residues (Asp, Glu, Arg, Lys)	May bind charged or polar ligands; salt bridge potential
Catalytic residues (Ser, His, Asp triad)	Active site of a serine protease or hydrolase
Cysteine near the pocket	Potential covalent binding site

PyMOL visualization

When visualization is enabled, the output includes a downloadable PyMOL script:

Pocket surfaces: Each predicted pocket rendered as a colored surface
Residue sticks: Pocket-lining residues displayed as stick representations
Color coding: Pockets colored by rank (green = top, yellow/orange = secondary)
Labels: Pocket rank and score labels in the 3D view
Usage: Open in PyMOL with @p2rank_results.pml

Model-specific output notes

Default model

Standard feature set; results are directly comparable across experimental structures
Best accuracy on high-resolution crystal structures

AlphaFold model

Accounts for prediction uncertainty in pocket detection
P2Rank does not read pLDDT: the alphafold model simply ignores the B-factor column, where predicted structures store pLDDT. Check the pLDDT of pocket-lining residues yourself
May produce fewer pockets on predicted structures compared to experimental structures

Downloading artifacts

CSV export: Pocket table with scores, druggability, residues, and all properties
PDB file: Input structure with pocket occupancy encoded in the B-factor column
PyMOL script: Interactive visualization script for offline exploration
JSON export: Full structured results with metadata and model information

Canvas visualization

When a P2Rank job completes, a result card appears on the canvas.

Card size: 450x560px
Border color: Green (#4CAF50)
Visualization: 3D Mol* viewer with highlighted binding pockets on protein surface

Tier 2 metrics

Pocket count
Top pocket druggability score
Residue count per pocket
Model type used (default, alphafold)

Interpreting canvas metrics

Metric	Good Value	Concern
Top pocket score > 0.8	High-confidence binding site	--
Top pocket score 0.5-0.8	Probable binding site	Validate with known data
Top pocket score < 0.5	Uncertain binding site	May need better structure
Druggable classification	Target is amenable to small molecules	--
Undruggable classification	Target may require alternative approaches	Consider fragments, PPI inhibitors

Quality checks

Verify that the top-scoring pocket corresponds to the known active site (if available for validation)
Check that pocket residues include expected catalytic or binding residues
Check that the pocket's lining residues can plausibly hold the expected ligand (P2Rank reports no volume)
Compare predictions across model variants if unsure about pocket quality
For predicted structures, cross-reference pocket residues with high-confidence (pLDDT > 70) regions

Interpreting P2Rank outputs