HIV-GRADE: Rules-Based Genotypic HIV Drug-Resistance Interpretation - - PowerPoint PPT Presentation

HIV-GRADE: Rules-Based Genotypic HIV Drug-Resistance Interpretation under Consideration of Bioinformatic Knowledge

Alejandro Pironti

Computational Biology and Applied Algorithmics Max Planck Institute for Informatics April 18, 2013

What is HIV-GRADE?

• HIV Genotypic Resistance Algorithm

DEutschland

• Rule set for drug-resistance interpretation
• Platform for comparing HIV-GRADE’s

interpretation with ANRS, HIVDB, Rega and geno2pheno[resistance]

• http://www.hiv-grade.de

April 18, 2013 Alejandro Pironti

How are HIV-GRADE rules established?

April 18, 2013 Alejandro Pironti

Input

• Information extracted from

clinical studies

• Case reports from

cooperating institutions

• geno2pheno[resistance]

weights for each drug Discussion by HIV- GRADE Members Rule Proposal Voting on Rules Output: Consensus Rule Set

geno2pheno[resistance]

• Linear Support Vector

Regression

• Dataset: Up to 968

genotype-phenotype pairs for each drug

• Raw prediction output:

Resistance Factor (RF)

• Z-score normalizes RF

w.r.t. sequences from therapy-naïve patients

April 18, 2013 Alejandro Pironti

Extracting Weights from geno2pheno[resistance]

• Linear Support Vector

Regression is

• regularized. To

estimate we solve E: Error Function λ: Regularization parameter

April 18, 2013 Alejandro Pironti

) ( β β + =

T

x x f

β λ

β

2 )) ( ( min arg

1

+ −

∑

= N i i i

x f y E

Very similar to linear regression you know Regularization term: Forces optimization to find β with smallest coefficient absolute values possible Positive weights increase predicted resistance Negative weights decrease predicted resistance

HIV-Grade Rules

• Weights in

geno2pheno[resistance] have a strong influence of HIV- GRADE’s rules

• AZT Rules:

– 151M → Resistant (R)

– ≥ 3 from {41L, 67N, 70R, 210W, 215Y, 219QE} → R – 2 from {41L, 67N, 70R, 210W, 215Y, 219QE} → intermediate – 2 from {41L, 67N, 70R, 210W, 215Y, 219QE} → limited susceptibility (LS)

• AZT_SP (Selective

Pressure) Rules (excerpt):

– 65R, 41L, 210W, 215YF → LS – 65R, 2 from {74V, 181C, 184VI}, 5 from {41L, 67N, 70R, 210W, 215FY, 219QE} and ≤ 2 from {41L, 210W, 215YF} → LS

April 18, 2013 Alejandro Pironti

Mutation with negative weight in geno2pheno Mutation with positive weight in geno2pheno

April 18, 2013 Alejandro Pironti

Acknowledgements

Max-Planck-Institut für Informatik

Thomas Lengauer Nico Pfeifer Joachim Büch

University of Düsseldorf

Björn Jensen

University of Cologne

Rolf Kaiser Mark Oette Saleta Sierra Aragon Finja Schweizer Elena Knops Maria Neumann-Fraune Eugen Schülter Eva Heger Claudia Müller

Medizinisches Labor Berg

Hauke Walter Martin Obermeier

Institut für Immunologie und Genetik Kaiserslautern

Martin Däumer Alexander Thielen