2.4 Sequence processing and data analysis
The raw reads were filtered through trimming and quality control steps
prior to taxonomic assignment according to the QIIME v.1.7.0 quality
control process (Caporaso et al. 2012). Adaptor/primer regions were
removed, and potential chimeras were removed using USEARCHv9.2 (Edgar
2013). Reads were clustered at 97% into Molecular Operational Taxonomic
Units (MOTUs) according to the standard setting in USEARCHv9.2 (Edgar
2013). High-quality clean reads that passed quality filtering were
queried against the full NCBI database using BLASTn according to
previous study (Berry et al. 2017). MOTUs were resolved to species,
genus, or higher, for COΙ animals or rbcL plants primer assays based on
the percent similarity threshold: Sequences with identity ≥ 99% to a
single species were considered as a “species match,” and as a “genus
match” if sequences had ≥ 98% similarity to one or more species within
the same genus. DNA sequences in this study were deposited into the NCBI
Sequence Read Archive (SRA) under accession number: PRJNA637184.
Alpha diversity (i.e., Chao1, Shannon and Simpson) matrices were
performed using QIIME and displayed using R v.3.3.3. software. To
evaluate the pattern of dispersion of samples within each season, beta
diversity was calculated with the euclidean distance. Beta diversity was
calculated using QIIME and visualized by two-dimensional principal
coordinate analysis (PCoA). Diversity was compared between different
seasons to assess temporal differences in diet composition. We also
compared the relative abundance of food items at various taxonomic
levels and at different seasons based on the linear discriminatory
analysis (LDA) effect size (LEfSe) method using LEfSe software.