Background:  “Codon volatility” has been proposed as a novel method to measure selection pressure on genes, using a single sequence. This approach has been applied to assess regional selection pressure in genomes of M. tuberculosis and P. falciparum. Here we apply this method to different HIV-coding regions of 1 sequence and compare it with established maximum likelihood- and distance-based comparative measurements of selection.

Methods:  Arbitrarily chosen full-length clade-B HIV-1 sequences were obtained from the LANL HIV Sequence Database. Selection was measured on the env, gag, and pol coding regions by 4 different methods:  codon volatility, PAML (Yang), approximate likelihood ratio at a site (Kosakovsky-Pond), and SNAP (Nei and Gojobori). Volatility scores of each codon were calculated based on the proportion of point-mutation neighbors encoding different amino acids. Maximum likelihood estimates of branch length and nucleotide substitution rate parameters were derived from entire alignments.

Results:  The volatility p values for gag, pol, and env of clade B isolate (AF538302) were .661, .418, and 213, respectively. Values of omega (dn/ds) as estimated by PAML were gag .211, pol .196, and env .623. Volatility based measurements of selection failed to correlate with estimates derived by all 3 comparative methods.

Conclusions:  Determining which genes and which sites within genes are under the greatest or least selective pressure will be important in rational vaccine design for HIV. Several methods for measuring selection are available; each approach has its strengths and limitations. Here we assess the utility of codon volatility as a tool to measure selection across the HIV genome. The volatility scores fail to correlate with estimates obtained by the maximum likelihood methods of Yang and Kosakovsky Pond, as well as global dn/ds ratios derived via sliding window analysis (SNAP). According to the codon volatility method, env appears to be under less selection pressure than pol or gag, in direct contrast with the results of all comparative methods as well as previously established concepts of immune selection on HIV. Deficits in the codon volatility approach may result from:  codon volatility being a reflection of local codon usage bias, rather than selective history; the influence of amino acid content on codon volatility; and the rapid erosion of genetic evidence of recent selection in genomes with high mutation rates.

Keywords: HIV; Selection Pressure; Codon Volatility