Algorithm and clinical validation M. Obermeier 4/2007 - PowerPoint PPT Presentation

Algorithm and clinical validation M. Obermeier 4/2007

Interpretation-systems � Free available: � Rule based: • ANRS • HIVdb • REGA � Truly bioinformatics based: • geno2pheno [resistance] • geno2pheno [THEO] M. Obermeier 4/2007

Interpretations-Systeme � commercial: � Rule based : • TrueGene HIV-1 • ViroSeq � Truly bioinformatics based: • Virtual phenotype M. Obermeier 4/2007

Yet another algorithm? � German initiative for standardization � predict clinical success not phenotype � Integration of combination therapies (resensitising effects) � Rule based systems can be faster adapted to new drugs and easier updated than bioinformatic approaches (need for data!) M. Obermeier 4/2007

HIV-GRADE base � Experts opinion � literature � genotype-phenotype correlations � genotype-virtual phenotype correlations (geno2pheno) � different databases consisting of treatment, genotype and clinical outcome � Scientific board meeting twice a year � actual algorithm version 04-2007 M. Obermeier 4/2007

Special characteristics of HIV-GRADE � Explicit results for resensitising effects _SP-nomenclature (selective pressure) � Results for boosted and non-boosted PIs � 5 level classification: � Hypersusceptible � Susceptible � Limited susceptibility � Intermediate resistance � Resistance M. Obermeier 4/2007

Basis of the HIV-GRADE internet-tool � HIV-Alg module from Stanford-HIVdb � PERL-source-code is freely available � Software is in development since 1999 � Algorithm Specification Interface (ASI) M. Obermeier 4/2007

Workflow sequences mutation-lists identify genes alignment on Consensus send sequences B Sequence to geno2pheno extraction of mutations rule-based analysis geno2pheno report detailled output batch output M. Obermeier 4/2007

Number of rules ANRS 89 REGA 82 HIVDB 18 scoring-rules (+60 comments) GRADE 303 M. Obermeier 4/2007

HIV-GRADE TDF Resistance Intermediate Limited susceptibility � 65R � 2 out of (41L, � 41L � 69 ins � 215F/Y 210W, 215 F/Y) � (41L or 210W) + 2 � 67N + 70R+ � 2 out of (67N, out of (67N, 70R, 219Q/E 70R, 219Q/E) � 70E � 151M 219Q/E) � 41L + 210W + 215 F/Y � 4 out of (41L, 67N, 70R, 210W, 215 F/Y, 219Q/E) M. Obermeier 4/2007

HIV-GRADE DRV Resistance Intermediate Limited susceptibility � 5 out of (11I, � 4 out of (11I, � 3 out of (11I, 32I, 33F, 47V, 32I, 33F, 47V, 32I, 33F, 47V, 50V(x2), 54M 50V(x2), 54M 50V(x2), 54M (x2), 54L, 73S, (x2), 54L, 73S, (x2), 54L, 73S, 76V (x2), 84V, 76V (x2), 84V, 76V (x2), 84V, 89V) 89V) 89V) M. Obermeier 4/2007

Algorithm Specification Interface (ASI) � Rule-based algorithms can be described using xml-syntax � Xml-Files available for Stanford- HIVdb, ANRS and REGA HIV-GRADE can be described in a compatible format. M. Obermeier 4/2007

Rule example <RULE> <CONDITION> EXCLUDE 65R AND (SELECT ATLEAST 2 AND NOTMORETHAN 2 FROM (74V,181C,184V)) AND (SELECT ATLEAST 5 FROM (41L,67N,70R,210W,215FY,219QE)) AND (SELECT ATLEAST 2 AND NOTMORETHAN 2 FROM (41L,210W,215YF)) </CONDITION> <ACTIONS> <LEVEL>5</LEVEL> </ACTIONS> </RULE> M. Obermeier 4/2007

Sequence entry form M. Obermeier 4/2007

Mutation list entry form M. Obermeier 4/2007

Internet Tool output common informations included sequences HIV-1 subtype all mutations resistance associated mutations scored mut results drugs all scored mutations comments M. Obermeier 4/2007

www.hiv-grade.de M. Obermeier 4/2007

Clinical Validation � n=365 � 5 centers � Active drug score from 0 (R) to 1 (S) (ADS) � Inclusion criteria � Treatment failure � Genotype � Treatment before and after change � VL 0-12 weeks before change � VL change 8-16 weeks after change in treatment

Treatment 300 250 before change 200 after change number 150 100 50 0 AZT ddC d4T 3TC TDF FTC NVP SQV RTV NFV APV ATV T20 ddI ABC DLV EFV IDV LPV_r TPV_r r drug M. Obermeier 4/2007

Active drug score (ADS) � transformation of a qualitative statement into a quantitative factor � Resistant => 0 � Intermediate => 0.33 � limited susceptibility => 0.66 � Susceptible => 1 � Hypersusceptible => 1.33 � Sum of all given drugs M. Obermeier 4/2007

Clinical validation GRADE True positive rate HIVDB REGA ANRS False positive rate M. Obermeier 4/2007

Simple linear regression • Simple model: VL = Δ + after _ change log( ) a * ADS b VL before _ change M. Obermeier 4/2007

simple linear regression HIV-GRADE R 2 =0.12 4 2 Δ LOGVL 0 -2 -1 0 1 2 3 Δ ADS M. Obermeier 4/2007

simple linear regression Stanford HIVdb R 2 =0.13 4 2 Δ LOGVL 0 -2 -2 -1 0 1 2 3 Δ ADS M. Obermeier 4/2007

simple linear regression ANRS R 2 =0.07 4 2 Δ LOGVL 0 -2 -2 -1 0 1 2 3 4 Δ ADS M. Obermeier 4/2007

simple linear regression REGA R 2 =0.13 4 2 Δ LOGVL 0 -2 -2 -1 0 1 2 3 Δ ADS M. Obermeier 4/2007

Multiple linear regression = ∑ # drugs VL Δ + after _ change log( ) a * ADS b drug drug ν ν VL = v 1 before _ change M. Obermeier 4/2007

Multiple linear regression results � In this cohort none of the algorithms can correctly predict the resistance against ABC. � HIVdb and GRADE are good in predicting APV resistance (whereas ANRS is not) M. Obermeier 4/2007

simple linear regression HIV-GRADE (w/o ABC) R 2 =0.17 4 3 2 Δ LOGVL 1 0 -1 -2 -1 0 1 2 3 Δ ADS M. Obermeier 4/2007

Multiple regression HIV-GRADE (w/o ABC) R 2 =0.28 4 3 Δ LOGVL 2 1 0 -1 -2 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 Δ ADS M. Obermeier 4/2007

Correction of ADS with VL before treatment change R 2 =0.36 4 2 Δ LOGVL 0 -2 1 2 3 4 5 Δ ADS VLco M. Obermeier 4/2007

HIV-GRADE association � Thomas Berg, Medizinisches Labor Dr. Berg, Berlin � Patrick Braun, PZB, Aachen � Martin Däumer, Institut für Virologie, Köln � Josef Eberle, Pettenkofer-Institut, München � Robert Ehret, PZB Aachen � Rolf Kaiser, Institut für Virologie, Köln � Nils Kleinkauf, Charité, Berlin � Klaus Korn, NRZ für Retroviren, Erlangen � Harm Müller, Fenner-Labor, Hamburg � Martin Stürmer, Institut für Medizinische Virologie, Frankfurt � Hauke Walter, NRZ für Retroviren, Erlangen