Stanford HIVdb Research Articles

Frequency and patterns of protease gene resistance mutations in HIV-infected patients treated with lopinavir/ritonavir as their first protease inhibitor.

Selection of protease mutations on antiretroviral therapy (ART) including a ritonavir-boosted protease inhibitor (PI) has been reported infrequently. Scarce data exist from long-term cohorts on resistance incidence or mutational patterns emerging to different PIs. We studied UK patients receiving lopinavir/ritonavir as their first PI, either while naive to ART or having previously received non-PI-based ART. Virological failure was defined as viral load ≥ 400 copies/mL after previous suppression <400 copies/mL, or failure to achieve <400 copies/mL during the first 6 months. pol sequences whilst failing lopinavir or within 30 days after stopping were analysed. Major and minor mutations (IAS-USA 2008-after exclusion of polymorphisms) were considered. Predicted susceptibility was determined using the Stanford HIVdb algorithm. Three thousand and fifty-six patients were followed for a median (IQR) of 14 (6-30) months, of whom 811 (27%) experienced virological failure. Of these, resistance test results were available on 291 (36%). One or more protease mutations were detected in 32 (11%) patients; the most frequent were I54V (n = 12), M46I (n = 11), V82A (n = 7) and L76V (n = 3). No association with viral subtype was evident. Many patients retained virus predicted to be susceptible to lopinavir (14, 44%), tipranavir (26, 81%) and darunavir (27, 84%). This study reflects the experience of patients in routine care. Selection of protease gene mutations by lopinavir/ritonavir occurred at a much higher rate than in clinical trials. The mutations observed showed only partial overlap with those previously identified by structural chemistry models, serial cell culture passage and genotype-phenotype analyses. There remained a low degree of predicted cross-resistance to other widely used PIs.

A non-infectious cell-based phenotypic assay for the assessment of HIV-1 susceptibility to protease inhibitors

HIV-1 genotyping is widely accepted as a diagnostic tool to optimize therapy changes in patients whose antiretroviral regimen is failing. Phenotyping can substantially complement the information obtained from genotyping, especially in the presence of complex mutational patterns. However, drug susceptibility tests are laborious and require biosafety facilities. We describe the molecular mechanism of a non-infectious HIV-1 protease phenotypic assay in eukaryotic cells and validate its applicability as a tool for monitoring drug resistance. A cloning vector containing the fusion protein green fluorescent protein-HIV-1 protease (GFP-PR) was modified to facilitate the insertion of HIV-1 protease from infected subjects. Real-time quantitative PCR and western blot analysis were used to establish the molecular mechanism of the new phenotypic assay. The method was validated by analysing HIV-1 protease from 46 clinical isolates. Statistical comparisons were made between values obtained using our assay and those reported from alternative standardized phenotypic assays. The capacity of HIV-1 protease to cleave cellular translation factors, such as the eukaryotic translation initiation factor 4 (eIF4GI) and the poly(A)-binding protein (PABP), led to cyclical accumulation of GFP that varied with the dose of protease inhibitors. Validation and comparison revealed a significant correlation with the Virco TYPE HIV-1 test (P < 0.0001, Spearman's ρ = 0.60), the Antivirogram test (P = 0.0001, Spearman's ρ = 0.60) and the Stanford HIVdb (P < 0.0001, Spearman's ρ = 0.69). This cell-based non-infectious phenotypic method with a well-understood molecular mechanism was highly reliable and comparable to other widely used assays. The method can be used for both phenotyping of HIV-1 viral isolates resistant to protease inhibitors and screening of new protease inhibitors.

Resolution of Discordant HIV-1 Protease Resistance Rankings Using Molecular Dynamics Simulations

The emergence of drug resistance is a major challenge for the effective treatment of HIV. In this article, we explore the application of atomistic molecular dynamics simulations to quantify the level of resistance of a patient-derived HIV-1 protease sequence to the inhibitor lopinavir. A comparative drug ranking methodology was developed to compare drug resistance rankings produced by the Stanford HIVdb, ANRS, and RegaDB clinical decision support systems. The methodology was used to identify a patient sequence for which the three rival online tools produced differing resistance rankings. Mutations at only three positions ( L10I , A71IV, and L90M ) influenced the resistance level assigned to the sequence. We use ensemble molecular dynamics simulations to elucidate the origin of these discrepancies and the mechanism of resistance. By simulating not only the full patient sequences but also systems containing the constituent mutations, we gain insight into why resistance estimates vary and the interactions between the various mutations. In the same way, we also gain valuable knowledge of the mechanistic causes of resistance. In particular, we identify changes in the relative conformation of the two beta sheets that form the protease dimer interface which suggest an explanation of the relative frequency of different amino acids observed in patients at residue 71.

Genetic variation of the HIV-1 integrase region in newly diagnosed anti-retroviral drug-naïve patients with HIV/AIDS in Korea

The survival time of HIV/AIDS patients in Korea has increased since HAART (highly active anti-retroviral therapy) was introduced. However, the occurrence of drug-resistant strains requires new anti-retroviral drugs, one of which, an integrase inhibitor (INI), was approved by the US Food and Drug Administration (FDA) in 2007. INIs have been used for therapy in many countries and are about to be employed in Korea. Therefore, it is important to identify basic mutant variants prior to the introduction of INIs in order to estimate their efficacy. To monitor potential drug-resistant INI mutations in Korean HIV/AIDS patients, the polymorphism of the int gene was investigated together with the pol gene using a genotypic assay for 75 randomly selected Korean HIV-1 patients newly diagnosed in 2007. The drug-resistant mutation sequences were analysed using the Stanford HIV DB and the International AIDS Society resistance testing-USA panel (IAS-USA). Seventy strains of Korean subtype B were compared with foreign subtype-B strains, and there were no significantly different variants of the int gene region in the study population. Major mutation sites in the integrase (E92Q, F121Y, G140A/S, Y143C/R, Q148H/R/K and N155H) were not detected, and only a few minor mutation sites (L74M, V151I, E157Q, V165I, I203M, S230N and D232N) were identified in 21 strains (28%). Resistance due to mutations in the pol gene was observed in a single strain (1.3%) resistant to protease inhibitors (PIs) and in four strains (5.3%) resistant to reverse transcriptase inhibitors (RTIs). In summary, this demonstrates that INIs will be susceptible to drug naïve HIV/AIDS patients in Korea.

Low rate of virological failure and maintenance of susceptibility to HIV‐1 protease inhibitors with first‐line lopinavir/ritonavir‐based antiretroviral treatment in clinical practice

Protease inhibitor (PI)-resistant HIV-1 has hardly ever been detected at failed boosted PI-based first-line antiretroviral regimens in clinical trials. However, this phenomenon has not been investigated in clinical practice. To address this gap, data from patients starting a first-line lopinavir/ritonavir (LPV/rtv)-based therapy with available baseline HIV-1 RNA load, a viral genotype and follow-up viral load after 3 and 6 months of treatment were extracted from the Italian Antiretroviral Resistance Cohort Analysis (ARCA) observational database. Based on survival analysis, 39 (7.1%) and 43 (7.8%) of the 548 examined patient cases had an HIV-1 RNA >500 and >50 copies/ml, respectively, after 6 months of treatment. Cox proportional hazard models detected baseline HIV-1 RNA (RH 1.79, 95%CI 1.10-2.92 per 1-log(10) increase, P=0.02) and resistance to the nucleoside backbone (RH 1.04, 95%CI 1.02-1.06 per 10-point increase using the Stanford HIVdb algorithm, P<0.001) as independent predictors of HIV-1 RNA at >500 copies/ml, but not at the >50 copies/ml cutoff criteria. Higher baseline viral load, older patient age, heterosexual route of infection and use of tenofovir/emtricitabine were predictors of failure at month 3 using the 50-copy and/or 500-copy threshold. Resistance to LPV/rtv did not occur or increase in any of the available 36 follow-up HIV-1 genotypes. Resistance to the nucleoside backbone (M184V) developed in four cases. Despite the likely differences in patient population and adherence, both the low rate of virological failure and the lack of development of LPV/rtv resistance documented in clinical trials are thus confirmed in clinical practice.

Comparison of HIV-1 Genotypic Resistance Test Interpretation Systems in Predicting Virological Outcomes Over Time

BackgroundSeveral decision support systems have been developed to interpret HIV-1 drug resistance genotyping results. This study compares the ability of the most commonly used systems (ANRS, Rega, and Stanford's HIVdb) to predict virological outcome at 12, 24, and 48 weeks.Methodology/Principal FindingsIncluded were 3763 treatment-change episodes (TCEs) for which a HIV-1 genotype was available at the time of changing treatment with at least one follow-up viral load measurement. Genotypic susceptibility scores for the active regimens were calculated using scores defined by each interpretation system. Using logistic regression, we determined the association between the genotypic susceptibility score and proportion of TCEs having an undetectable viral load (<50 copies/ml) at 12 (8–16) weeks (2152 TCEs), 24 (16–32) weeks (2570 TCEs), and 48 (44–52) weeks (1083 TCEs). The Area under the ROC curve was calculated using a 10-fold cross-validation to compare the different interpretation systems regarding the sensitivity and specificity for predicting undetectable viral load. The mean genotypic susceptibility score of the systems was slightly smaller for HIVdb, with 1.92±1.17, compared to Rega and ANRS, with 2.22±1.09 and 2.23±1.05, respectively. However, similar odds ratio's were found for the association between each-unit increase in genotypic susceptibility score and undetectable viral load at week 12; 1.6 [95% confidence interval 1.5–1.7] for HIVdb, 1.7 [1.5–1.8] for ANRS, and 1.7 [1.9–1.6] for Rega. Odds ratio's increased over time, but remained comparable (odds ratio's ranging between 1.9–2.1 at 24 weeks and 1.9–2.2 at 48 weeks). The Area under the curve of the ROC did not differ between the systems at all time points; p = 0.60 at week 12, p = 0.71 at week 24, and p = 0.97 at week 48.Conclusions/SignificanceThree commonly used HIV drug resistance interpretation systems ANRS, Rega and HIVdb predict virological response at 12, 24, and 48 weeks, after change of treatment to the same extent.

Evolving patterns of HIV‐1 transmitted drug resistance in Poland in the years 2000–2008

The aim of the study was to determine the rate of transmission of drug resistant human immunodeficiency virus-1 (HIV-1) variants among therapy-naïve HIV positive patients in Poland in the year 2008, to compare the data with the results from the years 2000 to 2007 and to monitor patterns of HIV-1 subtypes present in Polish population and their evolution. Complete protease and part of reverse transcriptase regions were sequenced from the sera of patients directed to the laboratory for drug resistance testing. The Stanford's HIVdb program was used for the interpretation of results and subtyping. The variants scoring at least "intermediate resistance" for at least one drug were considered as resistant. The results obtained were compared to those obtained in the years 2000-2007. A total of 95 patients were enrolled in the 2008 study. Homosexual transmission of infection was documented in more than 55% of all cases. The overall prevalence of transmitted drug resistance (TDR) was 5.3% (3.9% in 2007, 5.8% in 2006, and 14.1% in the years 2002-2005). The study from the years 2000 to 2001 revealed 28.7% prevalence. Preliminary analysis of the first half of 2009 shows the ratio of 7.8%. In four (4.2%) cases drug resistance was associated with protease inhibitors class, in one case (1.1%) with resistance to non-nucleoside reverse transcriptase inhibitors class. In four cases (4.2%) non-B subtype was identified (C, G, CRF01_AE, CRF02_AG). An increase of percentage of drug resistant mutants-from 3.9% (2007) to 5.3% (2008)-was recognized. In this study, TDR was limited to single classes of antiretroviral drugs. HIV-1 subtype B prevails in Poland.

Suboptimal Etravirine Activity is Common During Failure of Nevirapine-Based Combination Antiretroviral Therapy in a Cohort Infected with Non-B Subtype HIV-1

The primary objective of this study was to estimate etravirine activity in a cohort of patients infected with non-B subtype HIV-1 and failing nevirapine-based therapy. Genotypic resistance testing was performed if viral load was >OR= 1,000 copies/ml after receiving at least six months of therapy. Suboptimal response to etravirine was predicted by a score >OR= 2.5 on the Tibotec weighting schema, >OR= 4 in the Monogram schema, or classification as high to low-level resistant by a modification of the Stanford HIVdb algorithm (Version 5.1.2). Bivariate and multivariate analyses were conducted to determine the risk factors for suboptimal etravirine activity. The patients (n=91) were receiving nevirapine and lamivudine plus stavudine (57.1%) or zidovudine (42.9%). Median duration of nevirapine exposure was 53 weeks (IQR 46-101 weeks). The most common etravirine resistance associated mutations were Y181C (42.9%), G190A (25.3%), H221Y (19.8%), A98G (18.7%), K101E (16.5%), and V90I (12.1%). Suboptimal etravirine activity was predicted in 47.3 to 56.0%. There were disparities in mutations listed in Tibotec versus Monogram Schemas. Predicted suboptimal activity was not associated with nucleoside reverse transcriptase inhibitor (NRTI) used, gender, pretreatment or current CD4 cell count or viral load, subtype or NRTI mutations. Etravirine has compromised activity in approximately half of the patients failing nevirapine-based first-line treatment in this cohort, which supports guidelines that caution against using it with NRTIs alone in such patients.

Extracting causal relations on HIV drug resistance from literature

BackgroundIn HIV treatment it is critical to have up-to-date resistance data of applicable drugs since HIV has a very high rate of mutation. These data are made available through scientific publications and must be extracted manually by experts in order to be used by virologists and medical doctors. Therefore there is an urgent need for a tool that partially automates this process and is able to retrieve relations between drugs and virus mutations from literature.ResultsIn this work we present a novel method to extract and combine relationships between HIV drugs and mutations in viral genomes. Our extraction method is based on natural language processing (NLP) which produces grammatical relations and applies a set of rules to these relations. We applied our method to a relevant set of PubMed abstracts and obtained 2,434 extracted relations with an estimated performance of 84% for F-score. We then combined the extracted relations using logistic regression to generate resistance values for each <drug, mutation> pair. The results of this relation combination show more than 85% agreement with the Stanford HIVDB for the ten most frequently occurring mutations. The system is used in 5 hospitals from the Virolab project http://www.virolab.org to preselect the most relevant novel resistance data from literature and present those to virologists and medical doctors for further evaluation.ConclusionsThe proposed relation extraction and combination method has a good performance on extracting HIV drug resistance data. It can be used in large-scale relation extraction experiments. The developed methods can also be applied to extract other type of relations such as gene-protein, gene-disease, and disease-mutation.

Predicting the Response to Combination Antiretroviral Therapy: Retrospective Validation of geno2pheno‐THEO on a Large Clinical Database

Expert-based genotypic interpretation systems are standard methods for guiding treatment selection for patients infected with human immunodeficiency virus type 1. We previously introduced the software pipeline geno2pheno-THEO (g2p-THEO), which on the basis of viral sequence predicts the response to treatment with a combination of antiretroviral compounds by applying methods from statistical learning and the estimated potential of the virus to escape from drug pressure. We retrospectively validated the statistical model used by g2p-THEO in approximately 7600 independent treatment-sequence pairs extracted from the EuResist integrated database, ranging from 1990 to 2007. Results were compared with the 3 most widely used expert-based interpretation systems: Stanford HIVdb, ANRS, and Rega. The difference in receiver operating characteristic curves between g2p-THEO and expert-based approaches was significant (P < .001; paired Wilcoxon test). Indeed, at 80% specificity, g2p-THEO found 16.2%-19.8% more successful regimens than did the expert-based approaches. The increased performance of g2p-THEO was confirmed in a 2001-2007 data set from which most obsolete therapies had been removed. Finding drug combinations that increase the chances of therapeutic success is the main reason for using decision support systems. The present analysis of a large data set derived from clinical practice demonstrates that g2p-THEO solves this task significantly better than state-of-the-art expert-based systems. The tool is available at http://www.geno2pheno.org.

The Impact of Different Definitions on the Estimated Rate of Transmitted HIV Drug Resistance in the United Kingdom

The use of different lists of resistance mutations has resulted in estimates of transmitted HIV drug resistance (THDR) that are often not comparable. We estimated the rate of THDR based on the 3 definitions: > or =1 major mutation(s) listed on the International AIDS Society (IAS)-USA drug resistance mutation (DRM) 2006 list; > or =1 surveillance drug resistance mutation(s) (SDRM) on the published list by Shafer et al; and low-level/intermediate/high-level resistance to > or =1 drug(s) according to the Stanford HIVdb interpretation algorithm. Analyses were based on genotypic resistance tests conducted during 1997-2005 on antiretroviral therapy-naive patients and reported to the UK HIV Drug Resistance Database. The effect on THDR rates of revisions to the IAS-DRM list was also examined. Overall, 10.0%, 9.2%, and 10.4% of the 8272 samples available for analysis were classified as having THDR by the IAS-DRM, SDRM, and Stanford definitions, respectively; however, there was discordance for 244 (3%) samples. Changes in the version of the IAS-DRM list over time resulted in 4%-7% differences in the estimated rate of THDR, which increased from 4%-5% during 1997-2000 to 5%-7% during 2001-2005. The choice of genotypic definition had a minor influence on the estimated rate of THDR. The SDRM list is recommended for epidemiological estimates of THDR as it has been designed with such studies in mind.

Investigation of expert rule bases, logistic regression, and non-linear machine learning techniques for predicting response to antiretroviral treatment

The extreme flexibility of the HIV type-1 (HIV-1) genome makes it challenging to build the ideal antiretroviral treatment regimen. Interpretation of HIV-1 genotypic drug resistance is evolving from rule-based systems guided by expert opinion to data-driven engines developed through machine learning methods. The aim of the study was to investigate linear and non-linear statistical learning models for classifying short-term virological outcome of antiretroviral treatment. To optimize the model, different feature selection methods were considered. Robust extra-sample error estimation and different loss functions were used to assess model performance. The results were compared with widely used rule-based genotypic interpretation systems (Stanford HIVdb, Rega and ANRS). A set of 3,143 treatment change episodes were extracted from the EuResist database. The dataset included patient demographics, treatment history and viral genotypes. A logistic regression model using high order interaction variables performed better than rule-based genotypic interpretation systems (accuracy 75.63% versus 71.74-73.89%, area under the receiver operating characteristic curve [AUC] 0.76 versus 0.68-0.70) and was equivalent to a random forest model (accuracy 76.16%, AUC 0.77). However, when rule-based genotypic interpretation systems were coupled with additional patient attributes, and the combination was provided as input to the logistic regression model, the performance increased significantly, becoming comparable to the fully data-driven methods. Patient-derived supplementary features significantly improved the accuracy of the prediction of response to treatment, both with rule-based and data-driven interpretation systems. Fully data-driven models derived from large-scale data sources show promise as antiretroviral treatment decision support tools.

Drug-resistance mutations can be archived very early in HIV primary infection

Recent guidelines include clear recommendations for resistance testing in individuals infected by HIV for less than 2 years with the assumption that within 1--2 years most drug-resistance mutations (DRM) would be overgrown by wild-type virus. Recent studies addressed at monitoring the persistence of resistance-associated mutations after HIV transmission to untreated patients indicate that certain DRM are highly stable and may persist at detectable levels for many years. The usefulness of genotypic testing in all drug-naive patients in most settings is under debate. The picture of the events that occur in naive patients is probably complex and a different and more frequent analysis probably needs to be performed. (excerpt)

Analysis of discrepancies in the interpretation of antiretroviral drug resistance results in HIV-1 infected patients of Basque Country, Spain

Genotypic and phenotypic analysis of HIV-1 resistance mutations constitute one important point for providing guidelines in the choice of antiretroviral regimens and to design lines of rescue treatment for patients holding HIV-1 drug-resistant variants. However, some levels of discordance among them has been described. (i) To compare the genotypic analysis of resistance mutations to reverse transcriptase (RT) and protease (PR) inhibitors by Stanford HIVdb program (http://hivdb.stanford.edu) (St-HIVdb), and genotype with quantitative phenotypic analysis (Virtual Phenotype, VircoNET). (ii) To identify drug resistance mutations associated with discrepant results. Five hundred HIV-1 infected patients were included in this study. RNA was extracted from plasma. RT and PR regions were amplified and sequenced using ABI-Prism DNA sequencing system. Sequences were corrected and assembled with Seqman and Bioedit computer programs. The corrected sequences were submitted to the Stanford HIV-Seq program (http://hivdb.stanford.edu) and to Virtual Phenotype (VircoNET). Discrepant cases were considered if results were high or intermediate resistant by Stanford HIV-Seq program and susceptible by Virtual Phenotype, being detected as follows: (i) nucleoside RT inhibitors (NRT): 31.7% (ABC), 31% (d4T), 29.5% (ddC), 27.6% (ddI), 14.3% (TDF) and 11.3% (ZDV) and to PR inhibitors: 8.8% (SQV), 5% (APV), 3.8% (NFV) and 3.2% (IDV). These discrepant results were related to the presence of thymidine analogue mutations (TAMs) and also to key resistance mutations to NRT inhibitors: 65R, 69D/N, 74V/I, 184V/I and 215Y/F. (ii) PR inhibitors: 82A/F/T/S, 84I and 90M. Concordant results were considered when the interpretations by both programs were coincident, being higher than 96.7% for non-NRT inhibitors. The detection of discrepant results to NRT inhibitors and PR inhibitors, including the analysis of sequences with key resistant mutations to some drugs, means that further investigation is necessary in order to establish which is the best interpretation system as antiretroviral therapy guide.

Variability in the Interpretation of Transmitted Genotypic HIV-1 Drug Resistance and Prediction of Virological Outcomes of the Initial Haart by Distinct Systems

High level HIV-1 drug resistance in recently infected treatment-naive individuals correlates with sub-optimal virological responses to highly active antiretroviral therapy (HAART). To determine whether genotypic HIV-1 drug resistance in chronic naive patients, as interpreted by various systems, could predict the virological outcomes of HAART, isolates from patients enrolled in a prospective observational cohort (ICoNA) prior to treatment start were genotyped. Genotypic susceptibility scores (GSS) assigned to the initial HAART regimens using the interpretations of pre-therapy resistance mutations by 13 systems were related to virological outcomes. Of 415 patients, 42 (10%) had at least one major resistance mutation. According to the different interpretations, 1.9-20.5% of patients had some level of resistance to at least one drug in the initial regimen. In multivariable analysis, GSS from two systems significantly predicted the time to virological success: Rega 5.5, for each unit increase in GSS adjusted relative hazard (RH) 1.86 [95% confidence intervals (95% CI): 1.15-3.02] and hivresistanceWeb v3, RH 1.87 (95% CI: 1.00-3.48). With three other systems, GSS showed a trend towards a significant prediction of success: Retrogram 1.6, RH 2.33 (95% CI: 0.98-5.53), Menéndez 2002, RH 2.36 (95% CI: 0.97-5.72) and Stanford hivdb, RH 2.06 (95% CI: 0.94-4.49). Genotypic resistance testing coupled with adequate interpretation in chronic naive patients can usefully identify those at risk of sub-optimal virological response to HAART. This work was presented in part at the First European Workshop on HIV Drug Resistance. Luxembourg, 4-6 March 2003 (Abstract 44); and at the 9th European Conference on CIinical AIDS Therapy. Warsaw, 25-28 October 2003 (Abstract F6/7).

Comparative evaluation of three computerized algorithms for prediction of antiretroviral susceptibility from HIV type 1 genotype.

To compare three methods for using HIV-1 genotype to predict antiretroviral drug susceptibility. We applied three genotypic interpretation algorithms to 478 reverse transcriptase (RT) and 410 protease sequences for which phenotypic data were available. Sequences were obtained from clinical practice and from published sequences in the Stanford HIV-1 RT and Protease Sequence Database. The genotypic interpretation algorithms included: Stanford HIVdb program (HIVdb), the Visible Genetics/Bayer Diagnostics Guidelines 6.0 (VGI) and a genotypic interpretation program (AntiRetroScan, ARS) developed at the University of Siena, Italy. Genotypic interpretations were normalized to a three-level output: susceptible, intermediate and resistant. Discordances were defined as differences between genotype and phenotype for the same virus isolate. Discordances for which an isolate was considered susceptible by one test but resistant by another test were considered major discordances. The frequency of major discordances between genotype and phenotype was 10.6, 13.7 and 15.7% for ARS, VGI and HIVdb, respectively (P < 0.0001 for ARS versus HIVdb and for ARS versus VGI; P = 0.002 for VGI versus HIVdb). The correlation between genotype and phenotype was highest for non-nucleoside RT inhibitors and lowest for nucleoside RT inhibitors. Half of the major discordances involved stavudine, didanosine and zalcitabine. The concordance among the three genotypic algorithms was high, with weighted Kappa values ranging between 0.76 and 0.84 for the pairwise comparisons between each of the algorithms. Genotype interpretation algorithms correctly predict phenotype in 85-90% of cases, but the rate of concordance is not uniformly distributed among different drugs. These data provide insight into the potential additional benefit derived from phenotyping.

Construction, Training and Clinical Validation of An Interpretation System for Genotypic HIV-1 Drug Resistance Based on Fuzzy Rules Revised by Virological Outcomes

To evaluate whether fuzzy operators can be usefully applied to the interpretation of genotypic HIV-1 drug resistance by experts, and to improve the prediction of salvage therapy outcome by adapting interpretation rules of genotypic resistance on the basis of their association with virological response data. We used a clinical dataset of 231 patients failing highly active antiretroviral therapy (HAART) and starting salvage therapy with baseline resistance genotyping and virological outcomes after 3 and 6 months. A set of rules predicting genotypic resistance was initially derived from an expert (ADL). Rules were implemented using a fuzzy logic approach and the virological outcomes dataset used for the training phase. The resulting algorithm was validated using a separate set of 184 selected patients by correlating the resulting predicted activity with observed virological response at 3 months. For comparison, the expert systems from the drug resistance group of the Agence Nationale de Recherches sur le SIDA (ANRS-AC11) and the algorithm from the Stanford's HIV drug resistance database (Stanford HIVdb) were evaluated on the same set. The starting algorithm had a correlation with virological outcomes of R2=0.06 (P=0.0001). After the training phase the correlation with virological outcomes increased to R2=0.19 (P<0.000001). In the validation set of patients, the activity of the salvage regimen predicted by the fuzzy algorithm was the only variable independently predictive of the 3-month viral load change even after adjusting by the activity predicted by the two expert systems and baseline viral load (for each 10% salvage regimen's activity increase, mean HIV RNA change from baseline: -0.27 log10 copies/ml; 95% CI -0.39, -0.15). Using fuzzy operators in a virological outcomes training database to implement a rules-based algorithm for genotypic resistance interpretation, significant improvements of outcomes prediction were obtained. The resulting algorithm showed an independent predictive capability of virological outcomes over that of two rules-based interpretation algorithms made by experts. Although the system was trained and validated on a limited number of cases, the approach deserves further evaluation.