MotivationProtein solubility is an important property in industrial and therapeutic applications. Prediction is a challenge, despite a growing understanding of the relevant physicochemical properties.ResultsProtein–Sol is a web server for predicting protein solubility. Using available data for Escherichia coli protein solubility in a cell-free expression system, 35 sequence-based properties are calculated. Feature weights are determined from separation of low and high solubility subsets. The model returns a predicted solubility and an indication of the features which deviate most from average values. Two other properties are profiled in windowed calculation along the sequence: fold propensity, and net segment charge. The utility of these additional features is demonstrated with the example of thioredoxin.Availability and implementationThe Protein–Sol webserver is available at http://protein-sol.manchester.ac.uk.

Protein instability leads to reversible self-association and irreversible aggregation which is a major concern for developing new biopharmaceutical leads. Protein solution behaviour is dictated by the physicochemical properties of the protein and the solution. Optimising protein solutions through experimental screens and targeted protein engineering can be a difficult and time consuming process. Here, we describe development of the protein-sol web server, which was previously restricted to protein solubility prediction from amino acid sequence. Tools are presented for calculating and mapping patches of hydrophobicity and charge on the protein surface. In addition, predictions of folded state stability and net charge are displayed as a heatmap for a range of pH and ionic strength conditions. Tools are evaluated in the context of antibodies, their fragments and interactions. Surprisingly, antibody-antigen interfaces are, on average, at least as polar as Fab surfaces. This benchmarking process provides the user with thresholds with which to assess non-polar surface patches, and possible solubility implications, in proteins of interest. Stability heatmaps compare favourably with experimental data for CH2 and CH3 domains. Display and quantification of surface polarity and pH/ionic strength dependence will be useful generally for investigation of protein biophysics.

Predicting the concentrated solution behavior for monoclonal antibodies requires developing and using minimal models to describe their shape and interaction potential. Toward this end, the small-angle X-ray scattering (SAXS) profiles for a monoclonal antibody (COE-03) have been measured under solution conditions chosen to produce weak self-association. The experiments are complemented with molecular simulations of a three-bead antibody model with and without interbead attraction. The scattering profile is extracted directly from the molecular simulation to avoid using the decoupling approximation. We examine the ability of the three-bead model to capture features of the scattering profile and the dependence of compressibilty on protein concentration. The three-bead model is able to reproduce generic features of the experimental structure factor as a function of wave vector S(k) including a well-defined shoulder, which is a consequence of the planar structure of the antibody, and a well-defined minimum in S(k) at k ∼ 0.025 Å. We also show the decoupling approximation is incapable of accounting for highly anisotropic shapes. The best-fit parameters obtained from matching spherical models to simulated scattering profiles are protein concentration dependent, which limits their applicability for predicting thermodynamic properties. Nevertheless, the experimental compressibility curves can be accurately reproduced by an appropriate parametrization of the Baxter adhesive model, indicating the model provides a semiempirical equation of state for the antibody. The results provide insights into how equations of state can be improved for antibodies by accounting for their anisotropic shapes.

Improved understanding of properties that mediate protein solubility and resistance to aggregation are important for developing biopharmaceuticals, and more generally in biotechnology and synthetic biology. Recent acquisition of large datasets for antibody biophysical properties enables the search for predictive models. In this report, machine learning methods are used to derive models for 12 biophysical properties. A physicochemical perspective is maintained in analysing the models, leading to the observation that models cluster largely according to charge (cross-interaction measurements) and hydrophobicity (self-interaction methods). These two properties also overlap in some cases, for example in a new interpretation of variation in hydrophobic interaction chromatography. Since the models are developed from differences of antibody variable loops, the next stage is to extend models to more diverse protein sets. Availability The web application for the sequence-based algorithms are available on the protein-sol webserver, at https://protein-sol.manchester.ac.uk/abpred, with models and virtualisation software available at https://protein-sol.manchester.ac.uk/software.

In the present study, small-angle X-ray scattering (SAXS) and static light scattering (SLS) have been used to study the solution properties and self-interaction of recombinant human serum albumin (rHSA) molecules in three pharmaceutically relevant buffer systems. Measurements are carried out up to high protein concentrations and as a function of ionic strength by adding sodium chloride to probe the role of electrostatic interactions. The effective structure factors (Seff) as a function of the scattering vector magnitude q have been extracted from the scattering profiles and fit to the solution of the Ornstein–Zernike equation using a screened Yukawa potential to describe the double-layer force. Although only a limited q range is used, accurate fits required including an electrostatic repulsion element in the model at low ionic strength, while only a hard sphere model with a tunable diameter is necessary for fitting to high-ionic-strength data. The fit values of net charge agree with available data from potentiometric titrations. Osmotic compressibility data obtained by extrapolating the SAXS profiles or directly from SLS measurements has been fit to a 10-term virial expansion for hard spheres and an equation of state for hard biaxial ellipsoids. We show that modeling rHSA as an ellipsoid, rather than a sphere, provides a much more accurate fit for the thermodynamic data over the entire concentration range. Osmotic virial coefficient data, derived at low protein concentration, can be used to parameterize the model for predicting the behavior up to concentrations as high as 450 g/L. The findings are especially important for the biopharmaceutical sector, which require approaches for predicting concentrated protein solution behavior using minimal sample consumption.