We searched for molecular and functional consequences of ~12,000 coding variants, and were surprised that only a small fraction had strong effects on mRNA and protein expression, let alone on classic phenotypes. Is this resilience real or not? One obvious factor contributing to apparent phenotypic resilience may be inadequate depth of the phenome. Phenotypes may be detected only under specific conditions. For example, effects of the mutation in Nnt are much more pronounced under metabolic stress10. Artifactual resilience may also be caused by the presence of neighboring in-frame stop codons or splice acceptor or donor sites. For example, the loss of a stop codon in Dlgap5 only adds two amino acids due to the presence of a tandem stop codon six nucleotides downstream. Additionally, some negative results are likely to be caused by incorrect or incomplete gene models that generate spurious high impact variants.
