Limits...
Prediction of site-specific interactions in antibody-antigen complexes: the proABC method and server.

Olimpieri PP, Chailyan A, Tramontano A, Marcatili P - Bioinformatics (2013)

Bottom Line: They are extremely relevant as diagnostic, biotechnological and therapeutic tools.Their modular structure makes it easy to re-engineer them for specific purposes.We believe that it can be of great help in re-design experiments as well as a guide for molecular docking experiments.

View Article: PubMed Central - PubMed

Affiliation: Department of Physics, Sapienza University and Istituto Pasteur - Fondazione Cenci Bolognetti, 00185 Rome, Italy.

ABSTRACT

Motivation: Antibodies or immunoglobulins are proteins of paramount importance in the immune system. They are extremely relevant as diagnostic, biotechnological and therapeutic tools. Their modular structure makes it easy to re-engineer them for specific purposes. Short of undergoing a trial and error process, these experiments, as well as others, need to rely on an understanding of the specific determinants of the antibody binding mode.

Results: In this article, we present a method to identify, on the basis of the antibody sequence alone, which residues of an antibody directly interact with its cognate antigen. The method, based on the random forest automatic learning techniques, reaches a recall and specificity as high as 80% and is implemented as a free and easy-to-use server, named prediction of Antibody Contacts. We believe that it can be of great help in re-design experiments as well as a guide for molecular docking experiments. The results that we obtained also allowed us to dissect which features of the antibody sequence contribute most to the involvement of specific residues in binding to the antigen.

Availability: http://www.biocomputing.it/proABC.

Contact: anna.tramontano@uniroma1.it or paolo.marcatili@gmail.com

Supplementary information: Supplementary data are available at Bioinformatics online.

Show MeSH
Non-bonded contact prediction ROC curves for models A,B and C and the naive predictor
© Copyright Policy - creative-commons
Related In: Results  -  Collection

License 1 - License 2
getmorefigures.php?uid=PMC3753563&req=5

btt369-F1: Non-bonded contact prediction ROC curves for models A,B and C and the naive predictor

Mentions: We compared the results of the three predictors with those obtained using the naive predictor. The corresponding ROC curves are shown in Figure 1, and the MCC and the AUC values for all the models are shown in Table 3. In all cases, our models clearly outperform the naive predictor indicating that information on the whole antibody sequence effectively contributes to the prediction performance. All predictors had similar values of AUC and MCC for non-bonded contacts, whereas model B proved to be better at predicting both hydrogen bonds and hydrophobic interactions (See Supplementary Fig. S1). We obtained a slightly worse performance for main chain hydrophobic interactions and hydrogen bonds, possibly because of the small number of these interactions present in our dataset.Fig. 1.


Prediction of site-specific interactions in antibody-antigen complexes: the proABC method and server.

Olimpieri PP, Chailyan A, Tramontano A, Marcatili P - Bioinformatics (2013)

Non-bonded contact prediction ROC curves for models A,B and C and the naive predictor
© Copyright Policy - creative-commons
Related In: Results  -  Collection

License 1 - License 2
Show All Figures
getmorefigures.php?uid=PMC3753563&req=5

btt369-F1: Non-bonded contact prediction ROC curves for models A,B and C and the naive predictor
Mentions: We compared the results of the three predictors with those obtained using the naive predictor. The corresponding ROC curves are shown in Figure 1, and the MCC and the AUC values for all the models are shown in Table 3. In all cases, our models clearly outperform the naive predictor indicating that information on the whole antibody sequence effectively contributes to the prediction performance. All predictors had similar values of AUC and MCC for non-bonded contacts, whereas model B proved to be better at predicting both hydrogen bonds and hydrophobic interactions (See Supplementary Fig. S1). We obtained a slightly worse performance for main chain hydrophobic interactions and hydrogen bonds, possibly because of the small number of these interactions present in our dataset.Fig. 1.

Bottom Line: They are extremely relevant as diagnostic, biotechnological and therapeutic tools.Their modular structure makes it easy to re-engineer them for specific purposes.We believe that it can be of great help in re-design experiments as well as a guide for molecular docking experiments.

View Article: PubMed Central - PubMed

Affiliation: Department of Physics, Sapienza University and Istituto Pasteur - Fondazione Cenci Bolognetti, 00185 Rome, Italy.

ABSTRACT

Motivation: Antibodies or immunoglobulins are proteins of paramount importance in the immune system. They are extremely relevant as diagnostic, biotechnological and therapeutic tools. Their modular structure makes it easy to re-engineer them for specific purposes. Short of undergoing a trial and error process, these experiments, as well as others, need to rely on an understanding of the specific determinants of the antibody binding mode.

Results: In this article, we present a method to identify, on the basis of the antibody sequence alone, which residues of an antibody directly interact with its cognate antigen. The method, based on the random forest automatic learning techniques, reaches a recall and specificity as high as 80% and is implemented as a free and easy-to-use server, named prediction of Antibody Contacts. We believe that it can be of great help in re-design experiments as well as a guide for molecular docking experiments. The results that we obtained also allowed us to dissect which features of the antibody sequence contribute most to the involvement of specific residues in binding to the antigen.

Availability: http://www.biocomputing.it/proABC.

Contact: anna.tramontano@uniroma1.it or paolo.marcatili@gmail.com

Supplementary information: Supplementary data are available at Bioinformatics online.

Show MeSH