Quantifying Pathogen Surveillance Using Temporal Genomic Data

Chan, Joseph M.; Rabadan, Raul

These results indicate that H5N1 influenza virus surveillance of avian hosts is much more complete than H1N1 surveillance of swine. However, potential zoonotic transmission from other avian strains is possible, as well. Performance of the same analysis of non-H5N1 avian influenza virus strains yielded a low q2 coefficient beginning at 0.5 in 2003 and plateauing at a level just below 0.7 in spite of GISAID (Fig. 1D ) and the resulting increase in sequenced isolates (Fig. 1B) . This finding potentially reflects the extremely high genetic diversity of influenza A virus in its natural reservoir (Fig. 1C ) that is not fully captured by current surveillance systems. Transmission zone analysis of non-H5N1 avian influenza virus strains indicates that non-H5N1 surveillance is concentrated in southwestern Europe, the Central American-Caribbean region, North America, northern Africa, and eastern and southeastern Asia ( Fig. 2B ; see Fig. S1G in the supplemental material). Calculation of the q2 coefficient for other complete coding segments of influenza virus was also performed. Since observed differences in sequences should reflect the evolutionary rates of each segment, any differences in the q2 coefficient should reflect differences in sampling alone. For H3N2, H1N1pdm, and seasonal H1N1, the q2 coefficient exhibited little change among different segments (differences of Ͻ0.02) and moderate change for other strains. The largest change in the q2 coefficient was 0.186 between swine H1N1 HA and PB2. Despite such differences between segments, results showed that generally surveillance of human H3N2, H1N1pdm, and seasonal H1N1 across all segments surpassed that of human H5N1, avian H5N1, swine H1N1, and non-H5N1 avian influenza viruses (see Tables S2 and S3 in the supplemental material) .