2.1. Sequence processingIn this study, the sequences were processed using mothur, a soft-ware package with less computational demands [21]. Analysis of theraw data indicated that the reads covered V3 region successfully(size ranged ~ 200 bp). Forward and reverse reads were merged andN99% were overlapped at V3 region using the mothur pipeline (Refersupplementary Figs. S1, S2 and Table 2). The merged sequences werefurther processed. According to Huse et al. [22], accumulation of errorswithin a rather small subset of 454 readsmay occur hence itwas neces-sary to remove reads with ambiguous base calls (Ns), unusual or unex-pected length, low quality scores or those that cannot be aligned to thegene of interest (assumed to be unspecific PCR products) [22,23].Readswere trimmed based on quality scores, singletons (sequence reads thatoccur only once) are removed from the datasets to further reduce theerror rate [9].The mothur “seqNoise algorithm” incorporated with UCHIMEfurther removed chimeric sequences originated during PCR (5–45% ofPCR product) [24,25]. UCHIME was reported to perform best in a com-parative study where a reference databasewas used [26]. Critical analy-ses of different denoising tools demonstrated that parameters have tobe chosen very carefully so as not to introduce bias by readmodificationduring the generation of representative consensus reads.Hence,mothurwhich combined the above analyses such as OTU clustering, taxonomyassignment and multiple sample comparison, has been consideredto be more appropriate or the UPARSE pipeline [26,13,27] for OTUestimation. The resulting merged sequences and processing details areshown in Tables 2 and 3 and supplementary Table S1.
đang được dịch, vui lòng đợi..