Conservation of Hot Spots and Ligand Binding Sites in Protein Models by AlphaFold2

Abstract

The neural network-based program AlphaFold2 (AF2) provides high accuracy structure prediction for a large fraction of globular proteins. An important question is whether these models are accurate enough for reliably docking small ligands. Several recent papers and the results of CASP15 reveal that local conformational errors reduce the success rates of direct ligand docking. Here, we focus on the ability of the models to conserve the location of binding hot spots, regions on the protein surface that significantly contribute to the binding free energy of the protein-ligand interaction. Clusters of hot spots predict the location and even the druggability of binding sites, and hence are important for computational drug discovery. The hot spots are determined by protein mapping that is based on the distribution of small fragment-sized probes on the protein surface and is less sensitive to local conformation than docking. Mapping models taken from the AlphaFold Protein Structure Database show that identifying binding sites is more reliable than docking, but the success rates are still 5% to 10% lower than based on mapping X-ray structures. The drop in accuracy is particularly large for models of multidomain proteins. However, both the model binding sites and the mapping results can be substantially improved by generating AF2 models for the ligand binding domains of interest rather than the entire proteins and even more if using forced sampling with multiple initial seeds. The mapping of such models tends to reach the accuracy of results obtained by mapping the X-ray structures.

Figure 1.

a) Binding sites in surface representation for Entry 20 of the AlphaFold model P28720 (green), bound structure 1S39 (cyan), and unbound structure 4Q8M (red), with the ligand AQO inside each site, shown in sticks representation. b) Binding sites in surface representation for Entry 12 of the AlphaFold model P08709 (green), bound structure 5PAW (cyan), and unbound structure 1JBU (red), with the ligand 7XM inside each site, shown in sticks representation. c) Binding sites in surface representation for Entry 21 of the AlphaFold model P00734 (green), bound structure 3P70 (cyan), and unbound structure 2UUF (red), with the ligand BEN inside each site, shown in sticks representation.

Figure 2.

a) Complete AlphaFold model provided by the AlphaFold database for the protein with UniProt ID P00734. Contains multiple domains denoted by specific colors: orange - Gla domain, magenta - Kringle 1 domain, green - Kringle 2 domain, red - peptidase S1 domain, yellow - high-affinity receptor binding region. b) Binding sites of the X-ray structure 3P70_H (cyan), of the AF2 model downloaded from the AF database with UniProt ID P00734 (green), and of the AF2 model generated using only the sequence of the ligand-binding domain (yellow). Representative probes in the strongest hot spots predicted by FTMap are shown as wires in yellow, brown, and orange, respectively, for each model.

Figure 3.

a) All-atom RMSDs for the global alignment of the AF model P00734 from the AlphaFold database and the X-ray structures of 3P70_H with 90% sequence identity. 3P70_H is the PDB ID of the reference, ligand-bound X-ray structure and 2UUF_B is the unbound X-ray structure. RMSDs of both structures are depicted in red arrows on the graph. b) All-atom RMSDs for the global alignment of the AF2 model generated for the ligand binding domain of P00734 and the X-ray structures of 3P70_H with 90% sequence identity. 3P70_H is the PDB ID of the reference, ligand-bound X-ray structure and 2UUF_B is the unbound X-ray structure. RMSDs of both structures are depicted in red arrows on the graph. c) Density plots for the all-atom RMSDs of the global alignment of the AF models P00734 with X-ray structures of 3P70_H with 90% sequence identity.

Figure 4.

a) a) All-atom RMSDs from the local alignment of the AF model P00734 from the AlphaFold database and the X-ray structures of 3P70_H with 90% sequence identity. 3P70_H is the PDB ID of the reference, ligand-bound X-ray structure and 2UUF_B is the unbound X-ray structure. RMSDs of both structures are depicted in red arrows on the graph. b) All-atom RMSDs from the local alignment of the AF2 model generated for the ligand binding domain of P00734 and the X-ray structures of 3P70_H with 90% sequence identity. 3P70_H is the PDB ID of the reference, ligand-bound X-ray structure and 2UUF_B is the unbound X-ray structure. RMSDs of both structures are depicted in red arrows on the graph. c) Density plots for the all-atom RMSDs of the local alignment of the AF models of the P00734 with X-ray structures of 3P70_H with 90% sequence identity.