An investigation into the comparison of three human immunodeficiency virus (HIV) drug resistance interpretation algorithms

Human immunodeficiency virus (HIV) drug resistance is caused by mutations in the patient’s human immunodeficiency virus genome that renders antiretroviral (ARV) drugs less effective. Drug resistance not only results in the patient being more vulnerable to opportunistic infections, but also may increase the spread of resistant strains of HIV. Interpretation computer algorithms may be used to determine which ARV drug(s) the patient are resistant to, by analyzing the mutations that occurred in the patient’s HIV genome, instead of using expensive time consuming phenotypic laboratory tests. There are many different interpretation algorithms, but they often provide different resistance measures, even if applied to the same resistance profile. The aim of this study was to compare the latest versions of three HIV drug resistance interpretation algorithms in order to determine the extent of discrepancies between them. 2926 protease and 1981 reverse transcriptase subtype B sequences where obtained from the Stanford HIV-db genotype-phenotype correlation database. These sequences were pre-processed and the latest rules of the ANRS, HIV-db and REGA algorithms applied to them. The results were then compared with each other. These results indicate that although the accuracy of REGA, ANRS and HIV-db are similar, a deeper analysis of the results indicates that the interpretation algorithms are different. There need to be a mechanism of providing a single interpretation for a resistance profile of a genome. This may be created by collating the strengths of each of the interpretation algorithms.