Predicting specificity-determining residues in two large eukaryotic transcription factor families

Abstract

Certain amino acid residues in a protein, when mutated, change the protein's function. We present an improved method of finding these specificity-determining positions that uses all the protein sequence data available for a family of homologous proteins. We study in detail two families of eukaryotic transcription factors, basic leucine zippers and nuclear receptors, because of the large amount of sequences and experimental data available. These protein families also have a clear definition of functional specificity: DNA-binding specificity. We compare our results to three other methods, including the evolutionary trace algorithm and a method that depends on orthology relationships. All of the predictions are compared to the available mutational and crystallographic data. We find that our method provides superior predictions of the known specificity-determining residues and also predicts residue positions within these families that deserve further study for their roles in functional specificity.

Figure 1

Plot of score for basic leucine zipper DNA base specificity-determining residues versus the number of predictions considered. The vertical lines represent the number of the last prediction with a Z-score >3.0.

Figure 2

Plot of score for nuclear receptor DNA base specificity-determining residues versus the number of predictions considered. The vertical lines represent the number of the last prediction with a Z-score >3.0.

Figure 3

Plot of giant component size and score for basic leucine zipper DNA base specificity-determining residues predicted using the top eight predictions for all possible different sequence identity cutoffs.

Figure 4

Plot of giant component size and score for nuclear receptor DNA base specificity-determining residues predicted using the top ten predictions for all possible different sequence identity cutoffs.

Figure 5

Crystal structure of GCN4 bound to DNA (pdb:1YSA) (62). The eight predicted residues for the basic leucine zipper family are shown in VDW representation. The top ranked residue is colored red and the eighth is colored blue, with the colors shifted stepwise from red to blue for residues two through seven. The protein is shown in ribbon representation (colored black) and the DNA in lines representation (black). Figure made using VMD (63).

Figure 6

Crystal structure of GCR bound to DNA (pdb:1R4R) (49). The ten predicted residues for the nuclear receptor family are shown in VDW representation. The top ranked residue is colored red and the tenth is colored blue, with the colors shifted stepwise from red to blue for residues two through nine. The protein is shown in ribbon representation (colored black) and the DNA in lines representation (black). Figure made using VMD (63).