Table 2.

Summary for peptide–protein interactions docking methods.

Methods	Key Features	Model Quality #	Web Server	Refs
Local Docking
DynaDock	▪Combined Optimized Potential Molecular dynamics (OPMD) with a soft-core potential ▪Faster conformational sampling ▪Smoothened van der Waals and Coulomb energy potentials ▪Full flexibility of peptide and target protein	Near-native	Not available to public	[92]
Rosetta FlexPepDock	▪Monte Carlo-based optimization ▪High-quality conformational sampling ▪Hotspot residue (side-chain) modeling ▪Receptor flexibility (side-chains to full structure) ▪Rosetta energy function based clustering and scoring	Sub-angstrom *	http://flexpepdock.furmanlab.cs.huji.ac.il or http://www.rosettacommons.org/software	[93]
PepCrawler	▪Rapidly-exploring Random Tree (RRT) algorithm ▪Motion-planning based sampling ▪Ranking by automated energy funnel analysis (clustering-based) ▪Fully flexible peptide structures	Near-native *	http://bioinfo3d.cs.tau.ac.il/PepCrawler	[95]
Rosetta FlexPepDock ab initio	▪Ab initio modeling based on Rosetta fragment library ▪Simultaneous docking and de-novo folding of peptides ▪Peptide secondary structure option ▪No information for peptide conformation required	Near-native to Sub-angstrom §	http://www.rosettacommons.org/software	[96]
HADDOCK peptide docking	▪Modeling from ensemble of three canonical secondary structures (α-helix, extended or polyproline-II helix) ▪User-defined residues at binding pocket ▪Binding free energy based scoring ▪Fully flexible for interacting residues of peptide and protein	Near-native *	http://haddock.science.uu.nl/services/HADDOCK2.2/	[79,97]
PepSite 2.0	▪Identifies most peptide-binding site in seconds ▪Generates low-resolution model of peptide ▪Coarse-grained peptide orientation by spatial position-specific scoring matrix (S-PSSM)	Medium †	http://pepsite2.russelllab.org	[102]
Global Docking
ClusPro PeptiDock	▪Fast Fourier Transform (FFT)-based docking method ▪Motif-based prediction for peptide conformation ▪Clustering by structure scoring and CAPRI peptide docking criteria	Near-native to Sub-angstrom §	https://peptidock.cluspro.org/	[81,104]
pepATTRACT	▪Peptide structure prediction by threading sequence onto the three peptide conformations (as HADDOCK peptide docking) ▪Rigid-body peptide docking within binding pocket ▪Suitable for large-scale in silico protein–peptide docking ▪Clustering based on ATTRACT scores ▪Optional flexible docking for interacting residues	Near-native to Sub-angstrom §	http://bioserv.rpbs.univ-paris-diderot.fr/services/pepATTRACT/	[105]
HPEPDOCK	▪Hierarchical algorithm ▪Ensemble peptide conformation by MODPEP ▪Blind global peptide docking ▪Higher success rate and lower processing time for both global and local docking	Near-native to Sub-angstrom §	http://huanglab.phys.hust.edu.cn/hpepdock/	[109]
Template-based
GalaxyPepDock	▪Use similarity search (known template structures) as scaffolds for prediction ▪Energy-based model optimization and scoring ▪Superior accuracy using PeptiDB datasets than other servers	Medium (ligand); Near-native (interface)	http://galaxy.seoklab.org/pepdock	[110]
SPRINT-Str	▪Predict residues at peptide–protein binding interface ▪Use SVM with optimized parameters ▪Capability to distinguish binding sites of peptide from DNA, RNA and carbohydrate	N/A	http://sparks-lab.org/server/SPRINT-Str	[111]
PBRpredict-Suite	▪Predict interacting residues based on peptide-binding domain (PDB) from template sequences in NCBI database ▪Integrated six machine learning algorithms (model stacking) ▪Proteome-wide prediction feasibility	N/A	http://cs.uno.edu/~tamjid/Software/PBRpredict/pbrpredict-suite.zip	[112]
PepComposer	▪Motif similarity search to defined binding interfaces from monomeric protein databases PepX (http://pepx.switchlab.org) ▪Monte carlo-implemented PyRosetta ▪User-defined options for binding site residues or chain selection	Near-native	https://cassandra.med.uniroma1.it/pepcomposer/webserver/pepcomposer.php	[113]

^# RMSD of peptide backbone to experimental structure data. Medium: Between 2 Å to 5 Å; Near-native: 1 Å to 2 Å; Sub-angstrom: Less than 1 Å. * Tested on PeptiDB dataset. ^† Customized dataset of 405 known protein–peptide complexes with unbound receptor models. ^§ On selected subsets of PeptiDB.