Prediction of Bevi virimat Resist stance in HIV-1 Dr. Dominik - PowerPoint PPT Presentation

Arevir 2009 Prediction of Bevi virimat Resist stance in HIV-1 Dr. Dominik Heider Dept. of Bioinformatics, Center for Medical Biotechnology, April 2009 University of Duisburg-Essen dominik.heider@uni-due.de

2 Bevirimat • belongs to the class of maturation inhibitors • interfere with protease processing of precursor gag • gag is cleaved by the protease to produce functionally active proteins • unlike the protease inhibitors, Bevirimat binds the gag protein, not protease • prevents a critical cleavage at the p24-p2 junction • resulting virus particles lack functional capsid protein and have structural defects, rendering them incapable of infecting other cells. Dr. Dominik Heider

3 Bevirimat HIV-protease cleavage sites Modified from Salzwedel et al., 2009 Dr. Dominik Heider

4 p24 (part) p2 HIV-protease cleavage sites Dr. Dominik Heider

5 Classification of Bevirimat resistance • data set: • 45 susceptible/intermediate resistant sequences (fold change ≤ 10) • 110 resistant sequences • descriptors • hydrophobicity • molecular weight • solubility • HIV-protease cleavage site prediction • machine learning approaches • artificial neural networks • Random forests Dr. Dominik Heider

6 Results • NI-Rule • position 372 and 376 display a low variance, but different mean values for hydrophobicitiy • sequences with mutations in these positions can be filtered by calculating the differences of this position specific values and the mean values of either resistant and non-resistant sequences • sensitivity: 64% • specificity: 91% Dr. Dominik Heider

7 Results ROC-Plot • filtered sequences are passed to a neural network • best descriptor: hydrophobicity • sensitivity: 83.3% • Specificity: 100% • neural networks outperformed the Random Forests on filtered sequences Dr. Dominik Heider

8 Dr. Dominik Heider

9 Random Forests NI-Rule neural networks Dr. Dominik Heider

10 ID 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 wild type G H K A R V L A E A M S Q V T N S A T I M 0 0 0 0 0 0,68 1,75 0,11 0,64 0 1,78 0 0 0 0,06 0 0 0 0 0 0 8 S H K A R V L A E A M C Q A - N S T T V M 0 0 0 0 0 0,68 1,75 0 0,85 0 0,42 0 0 0 0 0,03 0,17 0 0 0 0 wild type G H K A R V L A E A M S Q V T N S A T I M H H H H H H H H H H H H H H H H 0 0 4 6 8 9 9 9 9 9 9 9 8 8 7 6 6 6 5 4 0 dQ369 G H K A R V L A E A M S V T N S A T I M H H H H H H H H H H H H H H H H 0 0 4 5 7 8 9 9 9 7 5 4 5 5 6 6 6 5 4 0 Dr. Dominik Heider

11 Conclusion • beside the QVT motif, the N and the I at position 372 and 376 respectively, play an important rule by determining the surface characteristics of the p24-p2 region • modification of the cleavage sites may lead to Bevirimat resistance • the accuracy of this prediction scheme is very high (84.8%), and therefore, can be used to preselect sequences for experimental studies Dr. Dominik Heider

12 People being involved... Daniel Hoffmann and Dominik Heider , Department of Bioinformatics, Center for Medical Biotechnology, University of Duisburg-Essen Jens Verheyen , Institute of Virology, University of Cologne Dr. Dominik Heider

13 Thank you very much for your attention Dr. Dominik Heider