To interrogate the sequence specificities of 104 mouse and 89 yeast transcription factors, Payne and Wagner mined data from protein-binding microarray studies, in which proteins are tested for their abilities to bind to all possible sequence combinations of short double-stranded DNA oligonucleotides. For each transcription factor, they assembled genotype–phenotype maps. In these network maps, each bound sequence is depicted by a node, and connections (that is, 'edges') between nodes represent one-step mutational routes (single-nucleotide substitutions, insertions or deletions) between different binding sites.
The genotype–phenotype maps revealed various interesting properties. Each map was typically large and highly interconnected (more so than expected by chance). This indicates robustness for transcription factor binding, as many binding-site mutations can be tolerated to maintain this in vitro phenotype of transcription factor binding. For example, for the mouse forkhead box protein A2 (FOXA2) transcription factor, bound sites can tolerate on average 37% of all possible single-nucleotide mutations. Additionally, the sequences with highest in vitro binding affinity showed the greatest robustness; these high-affinity sites were enriched for transcription factor binding in vivo from chromatin immunoprecipitation followed by sequencing (ChIP–seq) data, which implies that in vivo binding sites are frequently mutationally robust.
This is a preview of subscription content, access via your institution
