To establish the primary dataset of Arabidopsis (Arabidopsis thaliana) genes with a single mutant phenotype, we started with a published list of 620 Arabidopsis genes included in our sequence-based map of genes with mutant phenotypes (Meinke et al., 2003 (link)), removed problematic loci with questionable genotype-to-phenotype associations, eliminated suppressors and genes with a dominant gain-of-function mutant phenotype but no apparent loss-of-function phenotype, and further curated the phenotype and gene function information. Several classical genetic loci with well-characterized dominant phenotypes (e.g. GAI, ETR1, ABI1) were retained in the dataset because they are also associated with a distinctive loss-of-function phenotype (Peng et al., 1997 (link); Cancel and Larsen, 2002 (link); Ludwików et al., 2009 (link)). We then requested from Eva Huala at TAIR a list of genes that appeared to be associated with phenotype information in the TAIR database. Each locus on the list was evaluated. Most entries yielded useful information, but many candidate genes were eliminated because no suitable phenotype information was found or because the locus did not code for a protein. To complement these efforts, we initiated extensive PubMed searches of the scientific literature, using a combination of the following keywords: Arabidopsis, mutant(s), mutation(s), knockout, and null. Several thousand articles were retrieved and analyzed to obtain the information presented in Supplemental Table S2 . Information on genes with multiple mutant phenotypes was also retrieved with this approach. In order to proceed with further analysis of these datasets, no literature searches were performed for publications added to the PubMed database after December 31, 2010.
We then updated the information on essential genes based on the eighth release of the SeedGenes database (www.seedgenes.org ). Additional updates were obtained from a recent publication on embryo and gametophyte essentials of Arabidopsis (Muralla et al., 2011 (link)). We classified essential genes as being required for early development or survival. A locus was considered to be essential when knockout heterozygotes segregated for defective embryos or gametophytes, regardless of whether the resulting homozygotes remained viable to the seedling stage or beyond. To be consistent with the prioritized classification system established here, EMB loci with defects in gametophyte function were assigned to the gametophyte class instead of the seed/embryo class, regardless of whether the locus was classified elsewhere as being required for seed development, because that is when the mutant phenotype was first detected. The criteria used to differentiate between the GAM, GEM, EMG, and EMB subsets of essential genes are detailed elsewhere (Muralla et al., 2011 (link)). Mutants with defective gametes that produced viable homozygotes were typically assigned to the MGD subset. The criteria used to make other phenotype subset assignments, listed in Supplemental Table S1 , can be gauged by accessing the second tabbed spreadsheet in Supplemental Table S2 , sorting for the subset of interest, and evaluating the diversity of phenotypes represented.
We then updated the information on essential genes based on the eighth release of the SeedGenes database (