• RESISTANCE REPORTER© VOL 2, ISSUE 5 , Release: 2-Dec-2008
• Clinical Update 15, 2008 Euro HIV Drug Resistance Workshop
• Interactive Clinical Education Program, Don't miss your chance to participate!
• Clinical Update 14, 2008 International HIV Drug Resistance Workshop
• View the New Virco HIV Tool Kits
• View the New Interactive Learning Modules
• Laboratory Providers, watch a video on the new virco®TYPE HIV-1 Report
• Dr. Calvin Cohen Presentation on antiretroviral treatment regimens.
• Do not forget to "Opt-In" when you register!

• HOME
• HIV EDUCATION
• Interactive Clinical Education Programs
• Virco HIV Tool Kits
• HIV Clinical Updates
• HIV Patient Education
• HIV Learning Modules
• Archived HIV Presentations
• Laboratory Providers
• RESISTANCE REPORTER
• HIV/AIDS CONFERENCES
• RESISTANCE TEST ACCESS
• CONTACT US
• SITE MAP

Education > Clinical Updates

Clinical Update #11

Journal of Virological Methods
2007;145:47-55

The VirtualPhenotype^TM -LM is a data-driven linear-regression modeling engine that provides an accurate prediction of the HIV resistance phenotype from the viral genotype. In a previously reported study by Wang et.al., 2004 which compared linear regression modeling with six publicly available genotypic interpretation systems, this approach demonstrated the highest degree of accuracy in predicting HIV drug susceptibility. In the October 2007 issue of the Journal of Virological Methods, Hans Vermeiren and colleagues from Virco report on the use of an updated VirtualPhenotypeTM-LM system (version 4.0), which uses linear regression modeling. They found that the viral drug susceptibility can be accurately predicted from its genotype where the fold-change value is predicted as a sum of quantitative contributions of individual mutations and mutation pairs.

Quantitative levels of phenotypic resistance were calculated utilizing linear regression modeling (VirtualPhenotype^TM-LM v4.0). The datasets used for the linear regression models included samples with both a measured phenotype (Antivirogram^®) and a viral genotype on the same clinical isolate. The datasets ranged in size from approximately 6000 samples for recently approved drugs to approximately 40,000 samples for drugs that have been in use for a longer period of time. The linear models produced from this data include both mutations that are well recognized from publicly available mutations lists and newly identified mutations and pairs of mutations that contribute to a better prediction of drug susceptibility.

The linear regression models were assessed by the concordance of the predicted fold-change values with actual fold-changes reported by a measured phenotype (Antivirogram). Correlation was high except for DDI and D4T and to a lesser extent ABC and TDF. These drugs exhibit narrow dynamic ranges and the correlation was affected by assay variability. The correlation improved when assay variability was reduced by comparing the fold changes predicted by the models with the mean of multiple phenotypic measurements. The models were also evaluated by the ability of the LM fold-change values to predict virologic response when the drug was administered in a combination regimen. The clinical outcome data used in this analysis included two clinical cohorts and 13 clinical trials. Diagnostic accuracy by this measure was quite high for most drugs.

The authors conclude that linear modeling is a data driven approach for accurately predicting drug susceptibility. The models are frequently updated and have demonstrated the ability to identify the impact of new mutations on drug susceptibility that in some cases have not been reported by other sources. Because this approach is based on the effects of individual mutations and not on patterns, fold-change values can be predicted even for rare combinations that are not represented in the database.

If you are interested in viewing the complete journal article, please click here to link to download a copy for your own use.

PDF documents require Adobe ^® Reader for viewing.
Click here to download Adobe Reader for free.