Advisory Committee Chair
Lawrence J Delucas
Advisory Committee Members
Martha W Bidez
Christie G Brouillette
Richard A Gray
Yuhua Song
W William Wilson
Document Type
Dissertation
Date of Award
2014
Degree Name by School
Doctor of Philosophy (PhD) School of Engineering
Abstract
The first recombinant human protein drug, insulin expressed in e.coli cells, was approved by the FDA in 1982. Since then, protein therapeutics have become the fastest growing segment of the pharmaceutical industry and include immunoglobulin-g (IgG) directed cancer and immune disorder treatments. A major difficulty to bring protein drugs to market is the requirement that they be concentrated up to 150 mg/ml without aggregation for efficacy in a small injection volume. One way to improve protein drug solubility is to include additives that reduce protein-protein attraction and increase protein-protein repulsion thereby preventing protein molecules from coming together to form aggregates. However, hundreds of individual additives are approved by the FDA for injection and just ten additives at four possible concentration levels provides over a million (410) possible formulations. To address this formulation search problem, automated hardware and screening techniques are applied to evaluate the effect of additives on protein-protein interactions. These interactions are quantified by the second virial coefficient (B value), a thermodynamic parameter that is the sum of forces between two protein molecules at all orientations and distances in a solution. B values are measurements made by self-interaction chromatography. Contributions to the formulation search problem include hardware, software and screen methodology improvements. The hardware consists of 1) robotic formulation delivery and system equilibration, 2) a reduced-cost flow cell design with protein detection utilizing UV LEDs and 3) a multi-column system for parallel experimentation. The software includes two parts 1) automation of the self-interaction chromatography experiment and 2) a neural network model of additive influence on protein-protein interactions. The hardware and software components are utilized in a tiered additive screen including individual additive evaluation (initial screen), a complex formulation evaluation (incomplete-factorial) and training of a neural network to model the additive influence on protein-protein interactions. The neural-network model is then used to predict the B value of additive combinations not previously measured and the prediction capability of the model is evaluated. The system was evaluated using an IgG drug candidate protein from Minerva Biotechnologies and predictions from the neural network produced a 100 fold increase in solubility.
Recommended Citation
Johnson, David, "Development of a High-Throughput Self-Interaction Chromatography System" (2014). All ETDs from UAB. 2050.
https://digitalcommons.library.uab.edu/etd-collection/2050
