Thus most errors associated with 454 KRX-0401 sequencing should be corrected. Second, the long RT-PCR uses primers located in the conserved regions of HCV genome to allow the full recovery of viral variants. The inclusion of Deep Vent DNA polymerase in the enzyme blend not only facilitates the amplification of nearly full-length HCV genome but also reduces artificial mutations owing to its strong 39R 59 proofreading exonuclease activity. In our previous study, erroneous substitution rate is approximately at 0.13% after 60-cycle PCR cycles without the RT, translating into about 11.7 sites over 9022bp amplicon. Thus potential PCR-associated errors should have a minimal role on HCV mutation load measured in a genome-wide manner. Such a minimal role is further debilitated through a comparative analytical strategy in the present study. Third, individual patient HCV mutation loads didn’t correlate with viral titers, suggesting a minimal effect of the template amount on the mutation load. Finally, in the simulation experiment, the number of HVR1 structural variants and more importantly, the relative ratio of sequencing reads supporting each HVR1 structural variant, were fairly stable over various sequencing depths, indicating the lack of amplification bias at least at the level of structural variants. Indeed, in spite of the circulation as a heterogeneous population, the number of HCV HVR1 structural variants was limited, ranging 1 to 9 variants, and further reduced in terms of phylogenetic lineages. Taken together, while we are unable to map potential bias relevant to individual variants, above observation show that such a bias is unlike to have a role in the genome-wide mutation load that doesn’t have a focus on particular sites and mutations. Applying this method to well-characterized human subjects, we found the first unambiguous evidence that low mutation load in a genome-wide manner is associated with the better response to antiviral therapy. This conclusion remains effective when including only patients with unfavorable IL28B genotypes, one of the strongest single factors to predict treatment outcomes. Thus, the HCV genome-wide mutation load at high resolution appears to be an independent factor to predict the therapeutic results. While excess mutations are detrimental to viral population fitness, our data indicates a dominant role of natural selection even in patients with high mutation loads, which maximize the potential of HCV mutational pathways to counteract antiviral drugs, as observed in recent in vitro experiment. In-depth analysis of the mutation load explains previous conflicting data with regard to the predicting value of viral population structure in HCV antiviral therapy. The power-law pattern of mutation histogram had a low-bound point crossed with a mutation frequency at 17.5%, which is beyond the detection limit of common methods like gel shift analysis or cloning and Sanger sequencing. However, it should be noted that conventional cloning and Sanger sequencing is comparable with 454 sequencing for the recovery of structural HCV HVR1 variants at the level of phylogenetic lineages.