 |
|
|
| |
Several Genotyping Systems Offer Hope as Tropism Predictors
|
| |
| |
5th European HIV Drug Resistance Workshop
Cascais, Portugal
March 29, 2007
Mark Mascolini
Because phenotyping to determine HIV-1 coreceptor preference is expensive and time consuming, researchers have fashioned several genotype-based methods to infer viral tropism. Such systems are feasible because genetic determinants of viral tropism lie mainly within a 35 amino acid stretch on the V3 loop of HIV-1 gp120. At the Resistance Workshop, Spanish researchers found several of these genotype-based tools relatively reliable in determining viral tropism [1].
The research team compared seven genotypic tropism schemes, online at the following Web sites:
http://genomiac2.ucsd.edu:8080/wetcat/v3.html: C4.5, C4.5 with p8-p12, PART, SVM, and Charge Rule
http://coreceptor.bioinf.mpi-sb.mpg.de/cgi-bin/coreceptor.pl: geno2pheno coreceptor
http://ubik.microbiol.washington.edu/computing/pssm/: webPSSM
The study started with 40 viral sequences with known tropism from the US NIH AIDS Research and Reference Reagent Program. Twenty-eight of these sequences were R5 tropic, 4 of them X4 tropic, and 8 dual or mixed tropic. Predictive accuracy was 80% for webPSSM, 80% for SVM, 77.5% for geno2pheno coreceptor, 75% for C4.5, 75% for PART, 70% for C4.5 with p8-p12, and 70% for Charge Rule.
The webPSSM system proved 100% accurate for R5 and X4 isolates and read the mixed-tropic isolates as X4. Using webPSSM as the reference system, the investigators tested concordance of the other tools in determining tropism of virus from 272 people with HIV. PSSM called 92% of these sequences R5 tropic and the rest X4/R5. Concordance measured 92.3% with C4.5, 92.1% with C4.5 with p8-p12, 91.6% with Charge Rule, 85.96% with PART, 83.42% with geno2pheno, and 82.8% with SVM. Concordance was excellent on the R5-only viruses but hovered around 50% for the X4/R5 samples. Most genotyping methods overestimated rates of X4/R5 virus.
The authors concluded that "HIV tropism can be inferred accurately using V3 sequences by some but not all genotypic algorithms." The webPSSM system, they found, yielded the most accurate results.
Senior investigator Federico Garcia (University Hospital San Cecilio, Granada) stressed that the results do not mean genotype-based systems will be ready for clinical use when CCR5 antagonists get licensed. Workshop organizer Charles Boucher (University of Utrecht) noted that even phenotype-based methods like Monogram's Trofile assay have yet to be validated in terms of clinical response.
Reference
1. Chueca N, Alvarez M, Casanas M, et al. A comparison of 7 different tools
available to predict coreceptor usage based on V3 genetic sequences. 5th European HIV Drug Resistance Workshop. March 28-30, 2007. Cascais, Portugal. Abstract 53.
|
| |
|
|
|
|
|
