Limits...
Predictions of hot spot residues at protein-protein interfaces using support vector machines.

Lise S, Buchan D, Pontil M, Jones DT - PLoS ONE (2011)

Bottom Line: These are the amino acid types on which the original model did not perform well.We have therefore developed two additional SVM classifiers, specifically optimised for these cases.HSPred reaches an overall precision and recall respectively of 61% and 69%, which roughly corresponds to a 10% improvement.

View Article: PubMed Central - PubMed

Affiliation: Department of Computer Science, University College London, London, United Kingdom.

ABSTRACT
Protein-protein interactions are critically dependent on just a few 'hot spot' residues at the interface. Hot spots make a dominant contribution to the free energy of binding and they can disrupt the interaction if mutated to alanine. Here, we present HSPred, a support vector machine(SVM)-based method to predict hot spot residues, given the structure of a complex. HSPred represents an improvement over a previously described approach (Lise et al, BMC Bioinformatics 2009, 10:365). It achieves higher accuracy by treating separately predictions involving either an arginine or a glutamic acid residue. These are the amino acid types on which the original model did not perform well. We have therefore developed two additional SVM classifiers, specifically optimised for these cases. HSPred reaches an overall precision and recall respectively of 61% and 69%, which roughly corresponds to a 10% improvement. An implementation of the described method is available as a web server at http://bioinf.cs.ucl.ac.uk/hspred. It is free to non-commercial users.

Show MeSH

Related in: MedlinePlus

Ras/RalGDS complex.Mapping of HSPred predictions onto the the complex (PDB code: 1LFD). The monomers have been rotated to display the interface. Red residues are correctly predicted hot spots (true positives); blue residues are correctly predicted non hot spots (true negatives); yellow residues are non hot spots erroneously predicts as hot spots (false positives).
© Copyright Policy
Related In: Results  -  Collection


getmorefigures.php?uid=PMC3046169&req=5

pone-0016774-g003: Ras/RalGDS complex.Mapping of HSPred predictions onto the the complex (PDB code: 1LFD). The monomers have been rotated to display the interface. Red residues are correctly predicted hot spots (true positives); blue residues are correctly predicted non hot spots (true negatives); yellow residues are non hot spots erroneously predicts as hot spots (false positives).

Mentions: To further validate HSPred, we have applied it to the protein-protein complex Ras/RalGDS (PDB code: 1LFD). The Ras/RalGDS complex is not homologous to any of the complexes in the original data set and it can then be regarded as an independent external test case. Experimental values are available in [19], from which we have taken the data corresponding to interface alanine mutations ( on Ras and on RalGDS). HSPred correctly identifies hot spot (true positives) and non hot spot residues (true negatives). However, residues are wrongly predicted as hot spots (false positives). The predictions are illustrated in Figure 3. These results are in line with the cross-validated estimates in Table 1 and confirm the accuracy of HSPred.


Predictions of hot spot residues at protein-protein interfaces using support vector machines.

Lise S, Buchan D, Pontil M, Jones DT - PLoS ONE (2011)

Ras/RalGDS complex.Mapping of HSPred predictions onto the the complex (PDB code: 1LFD). The monomers have been rotated to display the interface. Red residues are correctly predicted hot spots (true positives); blue residues are correctly predicted non hot spots (true negatives); yellow residues are non hot spots erroneously predicts as hot spots (false positives).
© Copyright Policy
Related In: Results  -  Collection

Show All Figures
getmorefigures.php?uid=PMC3046169&req=5

pone-0016774-g003: Ras/RalGDS complex.Mapping of HSPred predictions onto the the complex (PDB code: 1LFD). The monomers have been rotated to display the interface. Red residues are correctly predicted hot spots (true positives); blue residues are correctly predicted non hot spots (true negatives); yellow residues are non hot spots erroneously predicts as hot spots (false positives).
Mentions: To further validate HSPred, we have applied it to the protein-protein complex Ras/RalGDS (PDB code: 1LFD). The Ras/RalGDS complex is not homologous to any of the complexes in the original data set and it can then be regarded as an independent external test case. Experimental values are available in [19], from which we have taken the data corresponding to interface alanine mutations ( on Ras and on RalGDS). HSPred correctly identifies hot spot (true positives) and non hot spot residues (true negatives). However, residues are wrongly predicted as hot spots (false positives). The predictions are illustrated in Figure 3. These results are in line with the cross-validated estimates in Table 1 and confirm the accuracy of HSPred.

Bottom Line: These are the amino acid types on which the original model did not perform well.We have therefore developed two additional SVM classifiers, specifically optimised for these cases.HSPred reaches an overall precision and recall respectively of 61% and 69%, which roughly corresponds to a 10% improvement.

View Article: PubMed Central - PubMed

Affiliation: Department of Computer Science, University College London, London, United Kingdom.

ABSTRACT
Protein-protein interactions are critically dependent on just a few 'hot spot' residues at the interface. Hot spots make a dominant contribution to the free energy of binding and they can disrupt the interaction if mutated to alanine. Here, we present HSPred, a support vector machine(SVM)-based method to predict hot spot residues, given the structure of a complex. HSPred represents an improvement over a previously described approach (Lise et al, BMC Bioinformatics 2009, 10:365). It achieves higher accuracy by treating separately predictions involving either an arginine or a glutamic acid residue. These are the amino acid types on which the original model did not perform well. We have therefore developed two additional SVM classifiers, specifically optimised for these cases. HSPred reaches an overall precision and recall respectively of 61% and 69%, which roughly corresponds to a 10% improvement. An implementation of the described method is available as a web server at http://bioinf.cs.ucl.ac.uk/hspred. It is free to non-commercial users.

Show MeSH
Related in: MedlinePlus