Abstract:Purpose. Demonstrate the ability of an artificial neural network (ANN), trained on a formulation screen of measured second virial coefficients to predict protein self-interactions for untested formulation conditions. Materials and Methods. Protein self-interactions, quantified by the second virial coefficient, B 22 , were measured by self-interaction chromatography (SIC). The B 22 values of lysozyme were measured for an incomplete factorial distribution of 81 formulation conditions of the screen components. Th… Show more
“…Despite the initial popularity of light scattering for determining second virial coefficients as a means of identifying conditions conducive to protein crystallization [8−10], the method has been largely supplanted by self-interaction chromatography [26][27][28][29][30][31][32][33][34]79] − a technique more compatible M A N U S C R I P T…”
, The osmotic second virial coefficient for protein self-interaction: use and misuse to describe thermodynamic nonideality, Analytical Biochemistry (2015),
“…Despite the initial popularity of light scattering for determining second virial coefficients as a means of identifying conditions conducive to protein crystallization [8−10], the method has been largely supplanted by self-interaction chromatography [26][27][28][29][30][31][32][33][34]79] − a technique more compatible M A N U S C R I P T…”
, The osmotic second virial coefficient for protein self-interaction: use and misuse to describe thermodynamic nonideality, Analytical Biochemistry (2015),
“…This is problematic because weak antibody self-and cross-interactions are often responsible for aggregation and polyreactivity, respectively. 6,7,12,[14][15][16][17][18][19] Nevertheless, numerous assays such as self-interaction chromatography (SIC) [20][21][22][23][24][25] and cross-interaction chromatography (CIC) [26][27][28] have been designed to identify these possibly troublesome antibodies early in the discovery program to avoid downstream issues. In these chromatography assays, increased retention of mAbs passing through a column conjugated with identical mAbs or a pool of polyclonal serum antibodies is indicative of attractive self-or cross-interactions, respectively.…”
“…This is an established method to train and prevent over-fitting of an ANN and is described in Bishop’s book, Neural Networks for Pattern Recognition [27]. In a previous publication [28], our lab compared this method to that of a standard general linear model (GLM) and found the ANN to exhibit reduced prediction error compared to the GLM.…”
Formulation development presents significant challenges with respect to protein therapeutics. One component of these challenges is to attain high protein solubility (> 50 mg/ml for immunoglobulins) with minimal aggregation. Protein-protein interactions contribute to aggregation and the integral sum of these interactions can be quantified by a thermodynamic parameter known as the osmotic second virial coefficient (B-value). The method presented here utilizes high-throughput measurement of B-values to identify the influence of additives on protein-protein interactions. The experiment design uses three tiers of screens to arrive at final solution conditions that improve protein solubility. The first screen identifies individual additives that reduce protein interactions. A second set of B-values are then measured for different combinations of these additives via an incomplete factorial screen. Results from the incomplete factorial screen are used to train an artificial neural network (ANN). The “trained” ANN enables predictions of B-values for more than 4,000 formulations that include additive combinations not previously experimentally measured. Validation steps are incorporated throughout the screening process to ensure that 1) the protein’s thermal and aggregation stability characteristics are not reduced and 2) the artificial neural network predictive model is accurate. The ability of this approach to reduce aggregation and increase solubility is demonstrated using an IgG protein supplied by Minerva Biotechnologies, Inc.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.