Multivariate analysis of the sequence dependence of asparagine deamidation rates in peptides

To develop a quantitative scheme to describe and predict asparagine deamidation in polypeptides using chemometric models employing reduced physicochemical property scales of amino acids.

Deamidation rates for 306 pentapeptides, Gly-(n-1)-Asn-(n+1)-Gly, with the residues n-1 and n+1 varying over the naturally occurring amino acids, were obtained from literature. A multivariate regression technique, called projection to latent structures (PLS), was used to establish mathematical relationships between the physicochemical properties and the deamidation half-lives of the amino acid sequences. Three reduced physicochemical property scales, amide hydrogen exchange rates (to describe the relative acidity of the amide protons) and flexibility parameters for the sequences were evaluated for their predictive capacity.

The most effective descriptors of the deamidation half-lives were reduced-property parameters for amino acids called zz-scores. The PLS models with the reduced property scales, combined with the hydrogen exchange rates and/or flexibility parameters, explained more than 95% of the sequence-dependent variation in the deamidation half-lives. The amide hydrogen exchange rate (i.e., amide proton acidity), hydrophilicity, polarizability, and size of amino acids in position n+1 were found to be the principal factors governing the rate of deamidation. The effect of amino acids in position n-1 was found to be negligible.

Chemometric analysis employing reduced physicochemical parameters can provide an accurate prediction of chemical instability in peptides and proteins. The relative importance of these various factors could also be determined.