Using machine‐learning‐driven approaches to boost hot‐spot's knowledge

Rosário‐Ferreira, Nícia; Bonvin, Alexandre M. J. J.; Moreira, Irina S.

doi:10.1002/wcms.1602

“…To enable large-scale detection of PPI-hot spots, high-throughput PPI-hot spot prediction methods have been developed. They generally fall into 2 categories: 14 (1) Methods that compute the binding energy/free energy difference between the wild-type protein and a mutant using classical force fields or empirical scoring functions. 11,[15][16][17][18][19][20][21][22][23] (2) Methods that employ classifiers such as nearest neighbor, support vector machines, decision trees, Bayesian/neural networks, random forest, and ensemble machine-learning models using various features including conservation, secondary structure, solvent-accessible surface area, and atom density.…”

Section: Introductionmentioning

confidence: 99%

PPI-hotspotID: A Method for Detecting Protein-Protein Interaction Hot Spots from the Free Protein Structure

Chen,

Sargsyan,

Wright

et al. 2024

Preprint

0

View full text Add to dashboard Cite

Experimental detection of residues critical for protein-protein interactions (PPI) is a time-consuming, costly, and labor-intensive process. Hence, high-throughput PPI-hot spot prediction methods have been developed, but they have been validated using relatively small datasets, which may compromise their predictive reliability. Here, we introduce PPI-hotspotID, a novel method for identifying PPI-hot spots using the free protein structure, and validated it on the largest collection of experimentally confirmed PPI-hot spots to date. We show that PPI-hotspotID outperformed FTMap and SPOTONE, the only available webservers for predicting PPI hotspots given free protein structures and sequences, respectively. When combined with the AlphaFold-Multimer-predicted interface residues, PPI-HotspotID, yielded better performance than either method alone. Furthermore, we experimentally verified the PPI-hot spots of eukaryotic elongation factor 2 predicted by PPI-hotspotID. Notably, PPI-hotspotID unveils PPI-hot spots that are not obvious from complex structures, which only reveal interface residues, thus overlooking PPI-hot spots in indirect contact with binding partners. Thus, PPI-hotspotID serves as a valuable tool for understanding the mechanisms of PPIs and facilitating the design of novel drugs targeting these interactions. A freely accessible web server is available at https://ppihotspotid.limlab.dnsalias.org/ and the source code for PPI-hotspotID at https://github.com/wrigjz/ppihotspotid/.

show abstract

“…11,[15][16][17][18][19][20][21][22][23] (2) Methods that employ classifiers such as nearest neighbor, support vector machines, decision trees, Bayesian/neural networks, random forest, and ensemble machine-learning models using various features including conservation, secondary structure, solvent-accessible surface area, and atom density. 14,[24][25][26][27][28][29][30][31][32][33][34][35][36] Most of the PPI-hot spot prediction methods rely on the protein complex structure and some are accessible via webservers; e.g., Hotpoint, 37 KFC2, 38 PredHS, 39 and PredHS2. 40 Fewer methods use only the free protein structure [41][42][43][44][45] or sequence, 34,[46][47][48][49][50][51][52][53][54] and SPOTONE (hot SPOTs ON protein complexes with Extremely randomized trees) is available as a web server.…”

Section: Introductionmentioning

confidence: 99%

PPI-hotspotID: A Method for Detecting Protein-Protein Interaction Hot Spots from the Free Protein Structure

Chen,

Sargsyan,

Wright

et al. 2024

Preprint

0

View full text Add to dashboard Cite

Experimental detection of residues critical for protein-protein interactions (PPI) is a time-consuming, costly, and labor-intensive process. Hence, high-throughput PPI-hot spot prediction methods have been developed, but they have been validated using relatively small datasets, which may compromise their predictive reliability. Here, we introduce PPI-hotspotID, a novel method for identifying PPI-hot spots using the free protein structure, and validated it on the largest collection of experimentally confirmed PPI-hot spots to date. We show that PPI-hotspotID outperformed FTMap and SPOTONE, the only available webservers for predicting PPI hotspots given free protein structures and sequences, respectively. When combined with the AlphaFold-Multimer-predicted interface residues, PPI-HotspotID, yielded better performance than either method alone. Furthermore, we experimentally verified the PPI-hot spots of eukaryotic elongation factor 2 predicted by PPI-hotspotID. Notably, PPI-hotspotID unveils PPI-hot spots that are not obvious from complex structures, which only reveal interface residues, thus overlooking PPI-hot spots in indirect contact with binding partners. Thus, PPI-hotspotID serves as a valuable tool for understanding the mechanisms of PPIs and facilitating the design of novel drugs targeting these interactions. A freely accessible web server is available at https://ppihotspotid.limlab.dnsalias.org/ and the source code for PPI-hotspotID at https://github.com/wrigjz/ppihotspotid/.

show abstract

“…Currently, there are various computational approaches devoted to the calculation of relative free energies (ΔΔ G ) upon mutation at protein–protein interfaces . They can be grouped into several categories, namely, empirical weighted functions, statistical and contact potentials, machine learning and alchemical methods, e.g., thermodynamic integration (TI), free energy perturbation (FEP), and nonequilibrium transition free-energy calculations based on Crooks fluctuation theorem (CFT) and Bennett acceptance ratio (BAR). − …”

Section: Introductionmentioning

confidence: 99%

Alchemical Calculation of Relative Free Energies for Charge-Changing Mutations at Protein–Protein Interfaces Considering Fixed and Variable Protonation States

Hernández González,

de Araujo

2023

J. Chem. Inf. Model.

1

0

View full text Add to dashboard Cite

The calculation of relative free energies (ΔΔG) for charge-changing mutations at protein–protein interfaces through alchemical methods remains challenging due to variations in the system’s net charge during charging steps, the possibility of mutated and contacting ionizable residues occurring in various protonation states, and undersampling issues. In this study, we present a set of strategies, collectively termed TIRST/TIRST-H+, to address some of these challenges. Our approaches combine thermodynamic integration (TI) with the prediction of pK a shifts to calculate ΔΔG values. Moreover, special sets of restraints are employed to keep the alchemically transformed molecules separated. The accuracy of the devised approaches was assessed on a large and diverse data set comprising 164 point mutations of charged residues (Asp, Glu, Lys, and Arg) to Ala at the protein–protein interfaces of complexes with known three-dimensional structures. Mean absolute and root-mean-square errors ranging from 1.38 to 1.66 and 1.89 to 2.44 kcal/mol, respectively, and Pearson correlation coefficients of ∼0.6 were obtained when testing the approaches on the selected data set using the GPU-TI module of Amber18 suite and the ff14SB force field. Furthermore, the inclusion of variable protonation states for the mutated acid residues improved the accuracy of the predicted ΔΔG values. Therefore, our results validate the use of TIRST/TIRST-H+ in prospective studies aimed at evaluating the impact of charge-changing mutations to Ala on the stability of protein–protein complexes.

show abstract

Using machine‐learning‐driven approaches to boost hot‐spot's knowledge

Cited by 10 publications

References 206 publications

PPI-hotspotID: A Method for Detecting Protein-Protein Interaction Hot Spots from the Free Protein Structure

PPI-hotspotID: A Method for Detecting Protein-Protein Interaction Hot Spots from the Free Protein Structure

PPI-hotspotID: A Method for Detecting Protein-Protein Interaction Hot Spots from the Free Protein Structure

Alchemical Calculation of Relative Free Energies for Charge-Changing Mutations at Protein–Protein Interfaces Considering Fixed and Variable Protonation States

Contact Info

Product

Resources

About