To improve variant calling, we here tailored the available BW2952 reference to our starting strain, compared different variant calling tools, refined our confidence criteria, and manually inspected unannotated variants (see Materials and Methods and the supplementary material, Supplementary Material online).
Nevertheless, when manually inspected, these variants could safely be identified.
After quality filtering, the sequence chromatograms were individually inspected and variants were manually reviewed by at least two researchers.
We uploaded the BAM files, the SNP position files and E. grandis gene annotation files (gff3) into IGV and visually inspected the variants from different genes to confirm the annotations.
By doing so, we avoid the unnecessary variant inference and inspect co-variants that would be neglected otherwise.
Variants stored in these files are rendered in the VCF view allowing the user to visually inspect the variant calls against the reference sequence and annotation (Fig. 2).
After this step, the reads overlapping each position are inspected to identify variant sites.
To investigate this possibility, we inspected wild-type and variant Gln1 variants for growth phenotypes.
In particular, we inspected known transcript variants.
The remaining two variants were visually inspected and determined as true variants.
The initial list of variant sites is then inspected by a machine-learning-based classifier that tries to separate variants likely to be polymorphic from those that might be calling artifacts (lists of known variants and common artifacts generated by the 1000 Genomes Project can often be used to train these classifiers) (4, 68, 71).
