Accelrys has added a new algorithm to the latest release of its Discovery Studio platform that is expected to be of interest to biotherapeutic developers.
The algorithm, which is the result of a four-year collaboration between the Massachusetts Institute of Technology and Novartis, helps protein engineers identify the size and location of regions on antibodies that are prone to aggregation and then predict mutations to improve the stability of biotherapeutic formulations.
The company licensed the algorithm from MIT and implemented it in Discovery Studio 3.1, the latest version of its modeling and simulation software that it released this week.
Accelrys expects computational biologists to be the primary users of the software with potential interest from biological formulation scientists, discovery scientists, immunologists, and biologists.
The company said that the algorithm is the first commercial release of an experimentally validated in silico method for predicting the risk and possible locations of protein-protein aggregation — a key issue in the development of biotherapeutics.
Unlike small-molecule drugs, which are pill-based, biological therapeutics such as antibodies are stored and administered at a very high concentration in solution. These solutions need to be stored in a stable form without the protein forming aggregates, denaturing, or degrading over time.
However, the "inherent instability of biopharmaceuticals to degradation processes such as protein aggregation, especially during long-term storage" is one of the “major problems encountered in developing biopharmaceutical therapies," Bernhardt Trout, a researchers in the department of chemical engineering at MIT and one of algorithm's developers, noted in a statement.
The spatial aggregation propensity code that Trout and his colleagues developed "helps scientists rank antibodies and other proteins for their propensity to aggregate and determine which sites govern this phenomenon,” he said. He added that this can be done “from early in the discovery stage through development with a view to selecting stable candidates and engineering stable biomolecules.”
The algorithm requires the user to input a three-dimensional structure of the antibody that will eventually be formulated, Trout told BioInform, pointing out that because many antibodies have similar sequences and structures, researchers can use homology modeling techniques to obtain a structure that resembles the likely end product.
The algorithm analyzes the structure for “degrees of hydrophobicity” at various local areas on the surface and pulls out regions that it considers important based on averaging parameters set by the user.
With the pertinent regions highlighted, a user can choose sites to mutate in order to increase the stability of the antibody, Trout said.
In addition, a user can use the "developability index," which incorporates local hydrophobicity and net charge to rank antibodies for their propensity to aggregate, he said.
An Early Look
Adrian Stevens, senior product marketing manager for life sciences at Accelrys, explained to BioInform that the company expects the new algorithm will help biotherapeutics developers identify at-risk proteins early in the discovery process. This would give formulation teams an opportunity to make adjustments to the protein — such as introducing selected mutations into the region — to make them more stable.
“What MIT has done is use the … experimental data to tune this prediction algorithm so that it can tell you the sites of aggregation on an antibody and it can rank them in order of their propensity to aggregate so you know where you need to start looking at doing mutational studies to try and design it out,” he said.
Biotherapeutics researchers predominantly use the hydophobicity of the protein’s surface to identify aggregation risk proteins and to a lesser extent its net charge, Stevens said, explaining that a lipophilic patch on the surface of a protein indicates an increased propensity to aggregate.
However, attempts to optimize synthetic antibodies for target antigens run the risk of increasing a protein’s propensity to aggregate irreversibly, degrade in solution, or trigger an immune response when administered, he said.
Typically, formulation groups run wet-lab experiments that can last for up to eight weeks to determine the aggregation risk for a given antibody. Alternatively, these groups use pre-developed formulas called platforms to stabilize antibodies.
But these platforms don’t work for all antibodies and a formulations team may have to create a special solution to deal with a “really badly behaved antibody.” These specialized formulations can cost up to $300,000 per day, per drug, Stevens said.
Having a predictive tool lets researchers “understand the aggregation risk as early as they can so that if you have got more than one antibody coming through for a particular pipeline, you can choose the best behaving one and hopefully align it to one of your formulation platforms effectively,” he said.
Furthermore, “if you know you've got a problem with aggregation,” the tool lets users “design it out if you can, or at least manage that propensity at an earlier stage in the process before it gets into the development.”
By including the aggregation code alongside Discovery Studio's broader design workflows, such as tools for protein-protein docking and prediction of thermal and mutational stability, the company believes the new capability will help users make decisions on what changes to make without risking the stability or efficacy of the antibody.
In addition, since Discovery Studio is built on the latest version of the company's Pipeline Pilot platform, users have a unified platform to bring together multiple tools and can parallelize calculations on compute grids and clusters.
“This is the benefit of going to the commercial environment because you get to answer all of these questions within the same product so that you can effectively optimize against multiple objectives at the same time,” Stevens said.
Novartis validated the algorithm for immunoglobulin G1 antibody formulations —the dominant monoclonal antibody in commercial use.
The validation process indicated that when hydrophobic patches were knocked out, the propensity to aggregation was substantially reduced, Trout said. In addition, his team created a “developability index” in which they ranked a set of antibodies and validated them against a set of molecules from Novartis.
However IgG1 isn’t the only antibody this algorithm is good for, Stevens said.
In fact, Trout’s team has used the algorithm for research on IgG2 and IgG4 as well as to identify specific motifs that lead to aggregation. He said that users would not have to make adjustments to get the code to work for other antibodies and proteins.
Along with licensing the code, Accelrys has made some enhancements to the command line scripts and initial code to speed up its performance. Specifically, it was able to reduce the time required for analysis of a full-length antibody from 24 hours to two minutes without a corresponding drop in accuracy.
Have topics you'd like to see covered in BioInform? Contact the editor at uthomas [at] genomeweb [.] com.