A Tool for Predicting the Success of First-Line Antiretroviral Therapies Alejandro Pironti Computational Biology and Applied Algorithmics Max Planck Institute for Informatics May 19, 2011
Motivation • The first antiretroviral regimen: – Best chance of sustained virological suppression – Selection under consideration of transmitted drug resistance – Simplicity and side effects also important for success Primary Drug Resistance Trends in the RESINA Cohort 20 Prevalence(%) 15 10 5 0 2001 2002 2003 2004 2005 2006 2007 2008 Year Alejandro Pironti May 19, 2011
Motivation • Available tools for assisting therapy selection: – Designed for therapy-experienced patients – Therapy-naïve patients: • Different mutations, e.g. T215Y → T215S • A subset of regimens Alejandro Pironti May 19, 2011
The Data • 2074 first-line antiretroviral therapies from EuResist and the RESINA cohort including: – Drug compounds Histogram of Drug Combinations Alejandro Pironti May 19, 2011
The Data • 2074 first-line antiretroviral therapies from EuResist and the RESINA cohort including: – PR and RT sequences – A viral load measurement for week 48 (37-59) after therapy start or an earlier one if therapy failed before Alejandro Pironti May 19, 2011
Dichotomization • Failure: 60 – VL > 50 cp/ml at week 48 50 (37-59) – Therapy stop before week 40 48 % 30 • Definition yielded – 1188 (57%) successes 20 – 886 (43%) failures 10 • Development set: 1854 0 therapies Virological Failure at Week 48 • Test set: 220 therapies Other Failure Success Alejandro Pironti May 19, 2011
Prediction Schematic Representation of • Linear support vector a Support Vector Machine machine was used to predict therapy success n i g r a M • Training with development set • Features: – Genotype – Drug combination – Interactions up to 3 rd order, e.g. mutation:mutation:drug Hyperplane • Feature selection: z-score ≥ 2 Alejandro Pironti May 19, 2011
Performance Evolutionary Engine (therapy-experienced Therapy-Naïve Tool tool) 10 CV Development Set (AUC=0.7739) Therapy-Experienced Set (AUC=0.7650) 5-fold cross-validation on Test Set (AUC=0.7647) Therapy-Naïve Set (AUC=0.4842) Test set. AUC = 0.7110 development set. AUC = 0.7135 True Positive Rate True Positive Rate False Positive Rate False Positive Rate Alejandro Pironti May 19, 2011
Selected Features • 438 out of 8090 features selected or included by default • Default features: – WHO Mutations for Surveillance of Transmitted Drug Resistance – Drug Combination • Selected features: – PR Mutations: 4S, 10I, 12K, 13V, 15L, 19I, 20R, 37D/T, 41K, 57K, 61D/H, 63Q/N/H, 64M, 67E, 74A, 82I – RT Mutations: 39A, 162A/G, 166I, 169D, 176S, 179D/I, 192N, 200A, 207G – 140 Drug-Mutation and 190 Drug-Drug Interactions Mutations for which also a literature reference could be found Alejandro Pironti May 19, 2011
Tool Development Status • Tool and web interface have been implemented • Tool will be part of geno2pheno[resistance] • Release scheduled together with geno2pheno[resistance]’s upgrade from 3.2 to 3.3. • http://www.geno2pheno.org Alejandro Pironti May 19, 2011
Example 1 • Patient with viral mutations: – PR: 3I, 37N, 57K, 63P, 65D, 77I – RT: 60I, 67N, 69D, 122K, RT: 60I, 65R, 67N, 69D, 122K, 135T, 178I/M, 214F, 245Q, 135T, 178M, 181C, 190S, 214F, 248D, 272A, 286A, 288S, 245Q, 248D, 272A, 286A, 288S, 296S, 297K, 311K/R 296S, 297K • FLART: 3TC+ABC+EFV • Reached a VL of 56 cp/ml after 4 months of therapy. However, 2 months later, his VL went up to 4.6 log. He was switched to 3TC+ABC+FPV/r after which his VL was promptly suppressed. Drug resistance mutations Alejandro Pironti May 19, 2011
May 19, 2011 Example 1 Alejandro Pironti
Example 2 • Patient with viral mutations: – PR: 3I, 37N, 39A, 41K, 60E, 63P, 71T, 77I, 93L – RT: 123E, 135V, 162C, 169D, 207E, 214F, 245M • FLART: 3TC+AZT+EFV • After 9 months of therapy the patient still had a residual viral load of 135 cp/ml. The patient’s therapy was switched to 3TC+ABC+EFV after which the patient’s viral load became undetectable. Alejandro Pironti May 19, 2011
Example 2 Sucessful, second-line ART Originally selected FLART Alejandro Pironti May 19, 2011
Outlook • Inclusion of Darunavir • Use of – Gag mutations – HLA types for prediction improvement Alejandro Pironti May 19, 2011
Acknowledgements University of Cologne Max-Planck-Institut für Informatik Rolf Kaiser Thomas Lengauer Marc Oette André Altmann Melanie Balduin Joachim Büch Saleta Sierra Aragon Alexander Thielen Finja Schweizer Elena Knops Maria Neumann-Fraune University of Erlangen Eugen Schülter Hauke Walter Eva Heger Claudia Müller Institut für Immunologie und University of Düsseldorf Genetik Kaiserslautern Björn Jensen Martin Däumer Alejandro Pironti May 19, 2011
Recommend
More recommend
