Fig. 5

The impact of reference bias on the quantification of gene expression in the accessions sequenced by He et al. [11]. A Estimated expression of the 60 genes identified as showing a lack of expression correlation by He et al. [11], using either the Chinese Spring RefSeq v1.1 transcriptome or the pantranscriptome reference as targets for kallisto pseudoalignment. The dashed black line represents x = y, which is the expected value if the reference is not affecting the estimation of gene expression. An accession lying on this dashed line has this gene’s expression estimated the same when using each reference. Red dots and green dots represent accessions in which a given gene has a TPM value < 50 or > 150%, respectively, when mapping to Chinese Spring than when mapping to the pantranscriptome reference. A red star indicates that in 25 or more accessions, the gene has an estimated expression less than half when mapping to Chinese Spring compared to when mapping to the pantranscriptome reference. A green star indicates that in 25 or more accessions, the gene has an estimated expression more than double when mapping to Chinese Spring compared to when mapping to the pantranscriptome reference. B Spearman’s correlation coefficient (SCC) between homoeologue pairs where one homoeologue is in the set of genes showing a lack of expression correlation identified by He et al. [11]. SCC scores were computed between AB, AD and BD homoeologue pairs and the lowest score was used. Triads in which any of the homoeologues were not present in the RefSeq v1.0 HC gene annotation were excluded. The significance of the difference between SCC scores when using the Chinese Spring reference compared to when using the pantranscriptome reference was calculated using a two-tailed t-test with no assumption of equal variance