Zwitterionic peptide self-assembly has been used in antifouling coating of medical devices , coating gold nanoparticles , , , biointerfaces, and peptide probed atomic force microscopy . The sequence and number of charged amino acids in zwitterionic peptides determine their structural properties , . Because amino acids that compose peptides are also found in nature, we can look for insight from nature. The surface of proteins is commonly populated by charged amino acids; they make up about 40% of human protein surfaces . The most common are lysine (E) and glutamic acid (K), which are capable of electrostatically interacting with water as well as each other. Their ionic hydration creates a strong solvation shell around them, as opposed to arginine (R) which interacts with E more than water , . Based on their strong hydration and other evidence, E and K provide a resistance of non-specific binding and that may be the reason for their common occurrence on protein surfaces . There are still open questions about the interactions between these charged groups, especially regarding the seemingly small chemical difference between aspartic acid (D) and E which has vastly different occurrences on protein surfaces. In this work, we simulate large multivalent zwitterionic peptide chains, (EK)6, (ED)6, (RK)6, and (RD)6, in explicit water using molecular dynamics with the enhanced sampling technique called parallel tempering well-tempered ensemble (PT-WTE) . We examined the interplay between salt bridging, hydration, peptide structure, and multivalent interactions. We used machine learning to isolate free energy basins and then compute a reaction network model (Markov state model) describing the kinetic and thermodynamic relationship between hydration and salt bridging. Our network model gives an interpretable understanding and reaction rates for the exchange of water, salt bridging, and the complexities of multiple interactions. Future work will introduce the complexity of monovalent and bivalent salts to see their role in these reaction network models.
 â E. R. L. Brisson, Z. Xiao, and L. A. Connal. Aust. J. Chem. 69.7 (2016), p. 705.
 â H. Ye et al. ACS Appl. Mater. Interfaces 7.40 (2015), pp. 22448â22457.
 â A. Parnsubsakul, S. Oaew, and W. Surareungchai. Nanoscale 10.12 (2018), pp. 5466â5473.
 â A. K. Nowinski et al. Langmuir 30.7 (2014), pp. 1864â1870.
 â R. Chang et al. Polym. J. 50.8 (2018), pp. 563â571
 â R. F. Sommese et al. Protein. Sci. 19.10 (2010), pp. 2001â2005.
 â A. K. Nowinski et al. J. Am. Chem. Soc. 134.13 (2012). PMID: 22401132, pp. 6000â6005.
 â A. D. White et al. Chem. Sci. 3.12 (2012), pp. 3488â3494.
 - M. Bonomi and M. Parrinello. Phys. Rev. Lett. 104.19 (2010), pp. 1â4.