All lipids were annotated with subsequent manual verification of MS/MS spectral matching for compound identifications (Supplementary Table 1). A total of 1,808 (SWATH) and 1,521 (DDA) lipids (Supplementary Table 2) were first integrated disregarding the acyl chain positions (sn1, sn2, sn3), double bond positions, and stereoisomers (E, Z). For example, TAG(16:0/16:1/16:2), TAG(16:0/16:2/16:1), TAG(16:2/16:1/16:0), TAG(16:2/16:0/16:1), TAG(16:1/16:0/16:2), and TAG(16:1/16:2/16:0) were considered the same lipid. Likewise, lysoPC 16:1(7Z) and lysoPC 16:1(9E) were regarded as the same. For the remaining 1,023 lipids, presence or absence in each of nine species was represented as a binary data matrix of size 1023×9 (Supplementary Data 2).
Hierarchical clustering analysis was performed using the R statistical language (http://www.R-project.org) and the package ‘amap’ (http://CRAN.R-project.org/package=amap). The distance was calculated by ‘correlation’ in the package. The linkage was performed by ‘average’. We cited the previous report27 as the standard taxonomic tree.