Protein-Protein Interaction (PPI) Model
Higher category : 【Bioinformatics】 Ligand-Receptor Interaction Database
1. Key Notes
2. Models
3. References
1. Key Notes
⑴ Binding affinities are commonly quantified by the dissociation constant (Kd) or inhibition constant (Ki).
⑵ General consideration
① General properties (e.g., atom types)
② Physico-chemical properties (e.g., excluded volume, partial charge, heavy-atom neighbors, hetero-atom neighbors, hybridization)
③ Pharmacophoric properties (e.g., hydrophobicity, aromaticity, H-bond donor/acceptor, ring member)
⑶ Datasets
① PDBbind database of version 2016
○ Subset 1. The general set contains all available data : 13,285 protein-ligand complexes
○ Subset 2. The refined set, the subset of the general set, contains 4,057 high-quality complexes in total
○ Subset 3. The core 2016 set : 290 complexes from the refined set and for a high-quality benchmark
③ CSAR-HiQ
○ CSAR-HiQ_51 : Derived from 176 protein-ligand complexes.
○ CSAR-HiQ_36 : Derived from 167 protein-ligand complexes.
④ Biolip
⑷ Protein-ligand interactions are common, but protein-protein interaction models are still relatively scarce.
2. Models
⑴ Binding position prediction model
① Example: AlphaFold multimer
② Generally, it is assumed that a ligand-receptor distance of less than 3Å will result in high binding affinity.
⑵ Binding affinity prediction model
① Class 1. Sequence-based method
○ Example: DeepDTA, DeepDTAF, DeepFusionDTA, GraphDTA, CAPLA
② Class 2. Structure-based method
○ Example: Pafnucy, OnionNet, FAST, IGN, IMCP-SF, GLI
Table. 1. The subclasses of structure-based binding affinity prediction models
③ Performance comparison
Table 2. Scoring performances of binding affinity models
3. References
⑶ Structure-based, deep-learning models for protein-ligand binding affinity prediction
Input : 2024.03.31 01:08