Enriching Peptide Libraries for Binding Affinity and Specificity Through Computationally Directed Library Design

Abstract

Peptide reagents with high affinity or specificity for their target protein interaction partner are of utility for many important applications. Optimization of peptide binding by screening large libraries is a proven and powerful approach. Libraries designed to be enriched in peptide sequences that are predicted to have desired affinity or specificity characteristics are more likely to yield success than random mutagenesis. We present a library optimization method in which the choice of amino acids to encode at each peptide position can be guided by available experimental data or structure-based predictions. We discuss how to use analysis of predicted library performance to inform rounds of library design. Finally, we include protocols for more complex library design procedures that consider the chemical diversity of the amino acids at each peptide position and optimize a library score based on a user-specified input model.

Keywords: Integer linear programming; Library design; Peptide engineering.

Figure 1

Flowchart of the library design process. The first two steps of gathering information from binding experiments or structure-based models and prioritizing substitutions will depend on the information available for the protein-peptide interaction of interest, so we present some general guidelines. The optimization of a degenerate codon library to encode the desired substitutions then proceeds in two parts: initial trimming of the codon choices based on codon size and score, followed by ILP library optimization to yield a library of a desired size with an optimal score. Finally, suggestions are given for analysis of the predicted behavior of the library based on input models, which can inform further rounds of library design to improve predicted library characteristics.

Figure 2

Output of the initial codon trimming step and the ILP library optimization step. (a) An example of the degenerate codon choices for one position as output by writeCodon.pl. The columns are labeled with their corresponding properties. (b) An example library design output by runILP.pl. The three columns are position, degenerate codon, and amino acids encoded. The total size in DNA sequences is under the limit set by the user (in this case 10⁷). The total size in protein sequences is the product of the number of amino acids encoded by each chosen codon. The score is the optimized value, the number of protein sequences composed entirely of preferred amino acids. The useful fraction is the product of the fraction of trinucleotides encoding preferred amino acids for each chosen codon.

Figure 3

Analysis of sequence scores for a library designed and then screened for specificity. Predicted KSBcl-2 binding is shown on the x-axis. Predicted binding to competitors Mcl-1 (a) or Bcl-x_L (b) are shown on the y-axis. A density plot of scores for the theoretical library is shown in gray scale. Scores for sequences from a library pool enriched for binding specificity to KSBcl-2 are overlaid in red. The blue points are for peptides that were tested in solution binding experiments and showed at least some margin of specificity for KSBcl-2 binding.