content in the sequences, especially in positions without a change in overall charge. This Danshensu (sodium salt) charge feature may be important for EZH2 inhibition and could be used to guide future inhibitor design. The designed peptide SQ026 from Run 3 also deserves some analysis as it is the only successful design where P30 and G33 mutations were tolerated and had the second lowest IC50 value after SQ037. Analyzing this sequence in reference to the other unsuccessful peptide designs from Run 3 and the successful designs from Run 4, a consistent mutation of S28N is observed for successful inhibitor design. This could be an important mutation for inhibition and perhaps is the reason that SQ026 tolerated mutations in P30 and G33, while SQ035 and SQ032 did not. This would also explain why Run 1 and Run 2 were less successful in their design. Not only does the more restrictive constraints not allow for the increase in charge constraints seen in Run 4, but since there are no asparagine residues in the initial sequences, there was no opportunity for the S28N mutation through simple rearrangement. This suggests that the constraints used were perhaps too restrictive and should be loosened in a similar manner to Run 3 and Run 4 for future designs. Overall, from this analysis it would seem that Run 4 contains the optimum set of constraints for this design, allowing for both increased charge content and the S28N mutations, while restricting changes to positions 30 and 33 that were largely unsuccessful. While the S28N mutation may be specific to EZH2 inhibitor design, the increased charge 442-51-3 distributor constraint may be characteristic of general histone-modifying enzyme inhibitor design and is worthy of further exploration. Analysis of the template-based constraints demonstrate how one can use the results from this study to guide future EZH2 and other histone-modifying enzyme design. In order to guide future peptide inhibitor design more generally, however, one must analyze the influence of the fold specific
