Identi fication of Pockets on Protein Surface to Predict Protein–Ligand Binding Sites - Identification of Ligand Binding Site and Protein-Protein Interaction Area - page 30

Input

Given protein structure

P 1 P 2 P i P N

A

Verify results from each predictor

Calculate the z-score value for each pocket
site in different predictors

Get the top 3 pocket sites from each
predictor

B

Clustering those 3 N pockets according to

their spatial similarity

×

Calculate the total z-score for each final
cluster

Rank the final clusters by their total z-score

Calculate the mass center for each cluster

Identify potential binding residues
(Residue mapping)

C

Prediction pocket sites (single point and clusters)
Prediction pocket sites from single methods
Potential ligand-binding residues for each pocket

Output

Fig. 2.3 The illustration of the workflow in metaPocket. Step A: calling each single method.
Step B: generating meta-pocket sites. Step C: mapping potential ligand-binding residues

Generating meta-pocket sites . After calling each method, metaPocket only
takes the first three pocket sites from each method into account. Thus, totally there
are 24 pocket sites and these pocket sites are somehow overlapped spatially. To
identify those overlapped pocket sites (we call them as “meta-pocket” sites), we use