Processing of Affymetrix Expression Dataset

Affymetrix gene expression values were normalized by the Robust multiarrays average algorithm using the software Robin (Lohse et al., 2010). Three samples (one replicate from leaf region 3 time point 1, one replicate from region 4 time point 1, and one replicate from region 4 time point 3) were excluded from further analysis based on atypical behavior in the Robin quality controls.

The dataset was filtered to retain only probesets where at least nine of 11 probes match, with maximum of two mismatches per maize gene model (genome release ZmB73_5b) and which match only to one gene model. In cases where two or more probesets matched to the same gene model, only the highest expressed was retained. Only 3583 passed these filters. For these probesets, a new Mapman (Thimm et al., 2004) based on Mercator (Lohse et al., 2014) mapping for the maize genome release Zm_B73_5b was assembled. In some cases (transcripts discussed in the text), annotation was manually checked and improved. Annotation for the 3583 retained probesets is contained in Supplemental Datafile S4.