Quantifying Pathogen Surveillance Using Temporal Genomic Data

Chan, Joseph M.; Rabadan, Raul

As a comparison to the q2 coefficient, we applied bar coding to determine the b 0 values of different influenza virus strains at filtration Hamming distances ranging from 0 to 200 (Fig. 4A) . This threshold is analogous to the R threshold of the q2 coefficient: 2 years of influenza virus evolution equivalent to 2 ϫ (5 ϫ 10 Ϫ3 ) ϭ 1% genetic distance. We therefore considered b 0 at a Hamming distance of 1% of the length of HA, or roughly 17 nucleotides, to be appropriate for comparison to the q2 coefficient (Fig. 4B) . For the most part, the q2 coefficient and b 0 are inversely correlated. For example, the low b 0 and high q2 coefficient of human H3N2, seasonal H1N1, and H1N1pdm indicate good surveillance. Non-H5N1 avian influenza virus has an extremely high b 0 and a low q2 coefficient. It should be noted that the non-H5N1 avian influenza virus data set contains 15 different HA subtypes, as opposed to all of the other single-subtype HA data sets considered. Thus, b 0 may need to be normalized by the expected number of HA subtypes. Nevertheless, b 0 for non-H5N1 avian influenza virus remains high even with normalization. However, as with p2, calculation of b 0 demands substantial computing power and time on the order of hours.