Principal component analysis (PCA) of gene expression and transcript expression

Relative relatedness and reproducibility of genes or transcripts among all samples were examined by PCA using GE and PSI, respectively. For PCA of relatedness in transcript‐level for ploidy, stage and tissue in each subgenome, PSI of PanAS events from diploid, tetraploid and hexaploid wheat cultivars were used. Expression data matrices were comprised of logarithm transformed TPM values including the genes containing PanAS events, which were used for PCA of gene expression in each subgenome. To include the subgenomes in PCA, only conserved AS triads were adopted. For further comparison of GE between species, we calculated differential expression based on a method we established in our previous work, called differential expression feature extraction (DEFE) method (Pan et al., ²⁰¹⁹; Xiang et al., ²⁰¹⁹). This method is based on negative binomial distribution model and considering transcript length in the meantime, which is ideal for the comparison across species with different transcriptome sizes. Based on the homoeologous AS triad list generated, PSI for each event triad from A, B and D genomes was extracted separately, and was then treated as the same element for the downstream PCA. PCA was performed using the PCA function in R FactoMineR package. Bi‐plots of PC1 & PC2 and PC3 & PC4 were plotted using R ggplot2 package, and variation explanation percentage for each principal component was calculated by the get_eigenvalue function in R FactoMineR package. 3D‐plot of PCA was generated using R plot3D package. PLS‐DA analysis was performed using R ropls package based on the same data set used for PCA and hierarchical clustering.