Gene ontology
For the IDGS dataset, we obtained a total of 490 candidate SNPs, of
which only four were found in coding genes and produced non-synonymous
substitutions (Table 3). Of these four, two were assigned to
uncharacterized proteins and both were found to be fixed for the
reference allele in NIDGS and variable in SIDGS. Considering the other
two candidate SNPs, one was found on the B3GNT8 gene, which codes for
the hexosyltransferase protein. This SNP was identified as a candidate
by the RDA, and its strongest association was with land cover type
“Grassland/Herbaceous” (0.384). The second missense candidate SNP was
found on the NPR1 gene, which codes for the guanylate cyclase protein.
It was identified as a candidate by both the pcadapt and RDA
analyses, and in the latter it was most strongly associated with “soil
particle size” (0.397). For details regarding the assignment of the
remaining candidate loci to additional variant classes, refer to Table
3. Considering the results for each analysis separately, thepcadapt , LFMM, RDA and pRDA identified 48, 52, 31 and 316
candidate loci, respectively, with 18, 13, 89 and 139 candidate SNPs
uniquely identified, respectively (Figure 4). Considering the different
categories considered, we considered 18, 472, 337 and 362 candidate loci
for the population structure outlier approach (pcadapt ), genotype
environment associations (‘GEA’), and including (‘POP’) and excluding
(‘noPOP’) population structure, respectively, of which 18, 26, 21 and 1
were unique to analysis type (Figure 4). Gene ontology enrichment
analysis of candidate SNPs found no evidence for enrichment of
particular cellular components, biological processes or molecular
functions for any of the individual or combined analyses considered.
For the NIDGS and SIDGS datasets, we found 150 and 18 candidate SNPs,
respectively. For NIDGS, two candidate SNPs were found in coding regions
and resulted in non-synonymous substitutions, although both were
assigned to uncharacterized proteins. For SIDGS, no SNPs were found to
result in non-synonymous substitutions (Table 3). For details regarding
the assignment of the remaining candidate loci to additional categories,
refer to Table 3. For these two datasets, we also found no significant
GO term enrichment for all identified candidate SNPs.