Large-scale evolutionary surveillance of the 2009 H1N1 influenza A virus using resequencing arrays

Lee, Charlie Wah Heng; Koh, Chee Wee; Chan, Yang Sun; Aw, Pauline Poh Kim; Loh, Kuan Hon; Han, Bing Ling; Thien, Pei Ling; Nai, Geraldine Yi Wen; Hibberd, Martin L.; Wong, Christopher W.; Sung, Wing-Kin

To build the likelihood model, we first determined the substitution bias on our platform by comparing the PM and MM probes (of both strands) of 25 028 true calls made by PBC from the two replicate microarray experiments of patient sample 380 mentioned in the previous section. For each true call, we generated a hybridization intensity reduction order by ranking the PM and MM probes of a particular strand in decreasing order of hybridization intensity and recording their respective frequencies (Table 1 ). Table 1 shows that for each PM probe encoding, certain hybridization intensity reduction orders occur much more frequently than others. For example, if the PM probe encoding is 'A' (regardless of strand), then it is most likely that the hybridization intensity reduction order is 'TGC' or 'GTC'. Thus, by matching the hybridization intensity reduction orders of its PM/ MM probes with that in Table 1 , we can compute the likelihood that the putative base call for a query base with ambiguous signals is correct (see 'Materials and Methods' section). In this way, we can recover base calls of ambiguous query bases exceeding a reasonably high likelihood threshold and achieve better accuracy and call rate than PBC.