bioinformatics / protein structure prediction · BEGINNER LESSON
PrankWeb
PrankWeb is a web server for predicting and visualizing ligand-binding sites in protein structures. It ranks likely binding pockets using the P2Rank algorithm, which estimates how likely a surface region is to bind a ligand.
What it is and what it measures
PrankWeb is an online tool used on protein structures, often provided as PDB files, to find places where a ligand may bind. A ligand is a molecule that can attach to a protein. The tool measures how likely each pocket on the protein surface is to be a binding site by giving scores and probabilities; higher values mean higher predicted confidence. It also identifies the amino acid residues that make up each pocket.
[1] Functional and Structure Prediction of Hypothetical Proteins From ...
How the prediction works
PrankWeb implements P2Rank, a machine-learning algorithm. Machine learning means the program has been trained to recognize patterns from examples rather than following only hand-written rules. P2Rank groups surface points into pockets using local geometry, hydrophobicity, and evolutionary conservation. Local geometry describes the shape of the surface, hydrophobicity describes how water-repelling a region is, and evolutionary conservation means positions that stay similar across related sequences, often found from multiple sequence alignments (MSAs). Together, these features help estimate where ligands are most likely to fit.
[1] Functional and Structure Prediction of Hypothetical Proteins From ...
How results are used
In the source example, protein structures predicted by AlphaFold3 were submitted to PrankWeb, which returned several putative pockets for each protein and ranked them by pocket score. The highest-ranked pocket is treated as the most promising candidate for follow-up study, but it is still only a prediction.
[1] Functional and Structure Prediction of Hypothetical Proteins From ...
Limitations
PrankWeb provides computational predictions, not direct proof that a pocket truly binds a ligand. The source explicitly notes that bioinformatics predictions need additional experimental validation. That means the result is a testable hypothesis, not final evidence of function.
[1] Functional and Structure Prediction of Hypothetical Proteins From ...
SEE THE IDEA
How PrankWeb produces a binding-site prediction
A 3D model or PDB file is submitted to the server.
The method inspects shape, water-repelling character, and conservation across related sequences.
Machine learning clusters surface points into pockets and assigns scores and probabilities.
Likely ligand-binding sites are listed from higher to lower confidence.
Input structure -> surface features -> pocket scoring -> ranked candidate binding sites
PrankWeb predicts where a ligand may bind by analyzing features of a protein surface and ranking the resulting pockets.
TRY IT YOURSELF
Test your understanding
Choose an answer, then check the explanation.
PUT IT TOGETHER
Match each concept to its meaning
Drag a label onto its description, or select a label and then select a description. This also works with a keyboard.
4 pairs to match
Sources and further reading
This general lesson was generated from public source material. It explains a concept; it does not interpret a private paper or provide medical advice.