Towards a Theory of Protein Adsorption: Predicting the Adsorption of Proteins on Surfaces via a Piecewise Linear Model

Predicting protein adsorption from solution to a surface is a perennial problem in biomedicine and related fields. Despite constant attention in the literature, it is not currently possible to predict quantitatively the amount of adsorbed protein given environment, protein and surface parameters. In previous work, we presented the Biomolecular Adsorption Database, an online collection of protein adsorption data collected from the literature, and more recently a program and set of algorithms for computing physico-chemical descriptors on protein surfaces. In this paper, we present a purely empirical approach to predicting protein adsorption using a linearly piecewise model with breakpoint. This model makes use of the previously developed surface property algorithms to describe the protein. We fitted and validated this model using the Biomolecular Adsorption Database. This model is capable of accounting for over 90% of the variance in the data, despite the fact that the adsorption data spans over three orders of magnitude. This represents a significant improvement over previous predictive modelling results.