Data from: The influence of population structure on gene expression and flowering time variation in the ubiquitous weed Capsella bursa-pastoris (Brassicaceae)

Population structure is a potential problem when testing for adaptive
phenotypic differences among populations. The observed phenotypic
differences among populations can simply be due to genetic drift, and if
the genetic distance between them is not considered, the differentiation
may be falsely interpreted as adaptive. Conversely, adaptive and
demographic processes might have been tightly associated and correcting
for the population structure may lead to false negatives. Here, we
evaluated this problem in the cosmopolitan weed Capsella bursa-pastoris.
We used RNA-Seq to analyse gene expression differences among 24
accessions, which belonged to a much larger group that had been previously
characterized for flowering time and circadian rhythm and were genotyped
using genotyping-by-sequencing (GBS) technique. We found that clustering
of accessions for gene expression retrieved the same three clusters that
were obtained with GBS data previously, namely Europe, the Middle East and
Asia. Moreover, the three groups were also differentiated for both
flowering time and circadian rhythm variation. Correction for population
genetic structure when analysing differential gene expression analysis
removed all differences among the three groups. This may suggest that most
differences are neutral and simply reflect population history. However,
geographical variation in flowering time and circadian rhythm indicated
that the distribution of adaptive traits might be confounded by population
structure. To bypass this confounding effect, we compared gene expression
differentiation between flowering ecotypes within the genetic groups.
Among the differentially expressed genes, FLOWERING LOCUS C was the
strongest candidate for local adaptation in regulation of flowering time.