Bioinformatics tools for coreceptor usage prediction for use of CCR5 antagonists
L.B. Arruda1, C.R. Gonsalez2, A.J.S. Duarte3, J. Casseb1
1Institute of Tropical Medicine/Unviversity of Sao Paulo, Laboratory of Investigation in Dermatology and Immunodeficiencies - LIM56, Sao Paulo, Brazil, 2Hospital of Clinics/School of Medicine/University of Sao Paulo, Out-clinic Ambulatory of Secondary Immunodeficiencies - ADEE3002, Sao Paulo, Brazil, 3School of Medicine/University of Sao Paulo, Laboratory of Investigation in Dermatology and Immunodeficiencies - LIM56, Sao Paulo, Brazil
Background: The clinical use of CCR5 antagonists requires
the coreceptor usage determination of viral strains from the infected
individual. The bioinformatics predictive programs could be a more accessible
alternative for the screening of candidates for the use of CCR5 antagonists
compared to the gold-standard assays to coreceptor usage determination.
Methods: DNA samples from 82
HIV infected subjects followed in our out-clinic (ADEE3002) were amplified by
PCR for the V3 loop region and sequenced. The coreceptor usage was performed
from the V3 loop sequences using nine predictive tests of bioinformatics
platforms (WetCat, WebPSSM and geno2pheno[coreceptor]) and the net charge of
the V3 sequences was calculated. Additionally, 79 V3 sequences with known
coreceptor usage from the Los Alamos data base
were tested using the same bioinformatics platforms aiming to evaluate the sensibility
and specificity of the tests.
Results: 49/82 (59.8%) of patients sequences were
discordant for at least one of the nine predictive tests. The assessing of the
net charge showed that 9/49 (18.4%) of these discordant results had net charges
< 4, suggesty these patients were not infected by X4 virus. Regarding the
concordant results (33/82; 40.2%), 29/33 (87.9%) were predicted as being R5
strains and 4/33 (12.1%) were predicted as being X4. The evaluation of the 79 Los Alamos sequences revealed that the nine
bioinformatics tests presented over 80% of sensitivity and specificity. The
concordance between known and predicted coreceptor usage was over 88% for all
tests used, and the SVM test (WetCat) had the best concordance rate (75/79; 95%).
Conclusions: Currently, the predictive systems for
coreceptor usage determination cannot be used alone; it is necessary to confirm
the results by a gold standard test. However, the prediction systems associated
with net charge determination could represent a more accessible strategy to
screening candidates to use the CCR5 antagonists.
Back to the Programme-at-a-Glance