Assays to Predict Developability
Antibody therapies are not novel. The first FDA approval and commercialization for an antibody therapeutic was in 1986 for Orthoclone OKT3, which was approved for prevention of kidney transplant rejection (Ecker et al. 2015). As of today, there are 78 monoclonal antibody treatments on the market to treat a wide variety of diseases, from arthritis to melanoma, and everything in between.
Predicting Effectiveness of Antibodies for Developability
Over the past 30+ years, the interest in antibodies has risen as scientists learn that robust and targeted treatments are within reach. For many years, chemists have relied on the Lipinski Rule of 5 to estimate the developability of a small molecule, but at this point, no such rule exists for large molecules. Unlike their small molecule cousins, antibodies are difficult to deliver and are even more difficult to characterize and optimize. Positing that patterns can be found to predict effectiveness of large molecules, Jain et al. (2017) set out to determine the necessary criteria. The team performed a meta analysis on data from 137 antibody therapeutics spanning twelve characterization assays. These antibodies were either in phase 2 or 3 of clinical trials or already FDA approved therapies. The approved therapies performed well while some of the drugs in the clinical trial phase had unfavorable results which may indicate developability risks. The results of the meta analysis seem to indicate that these twelve assays could provide a mechanism to measure the developability of an antibody. In his blog (Booth 2017), Bruce Booth reviewed this trend and coined this set of assays the “Dirty Dozen”.
Defining the Criteria and Leveraging the Platform for Science
In their paper, Jain et al (2017), leveraged R to plot and analyze their data using antibodies that are either approved or in clinical trials as their reference. To do this they set a criteria: antibodies that performed in the lowest 10% for a given assay received a flag. Once the analysis was complete, they totaled the number of flags for each antibody with the highest number of flags indicating the most unfavorable results. In doing this, they found that approved antibodies have fewer flags (65% had no flags), while a larger number of antibodies being tested in phase 2 clinical trials were flagged in one or more assays (only 40% had no flags). While no single assay could be successfully used as a predictor of developability, they reported that, taken together, the results of these twelve assays have the potential to help inform decisions in the antibody development pipeline. By combining the results of these assays, patterns started to emerge and new questions can be posed and tested.
As a company partnering with a variety of biotech companies, we have found that many scientists have the information they need to make decisions, but they lack the tools to bring that data together into something meaningful. The Platform for Science allows scientists to store all of their data in one location, perform calculations, and report on the results from a centralized tool. The Platform for Science Marketplace offers preconfigured applications to support a variety of needs, including tracking the samples, protocols, and data from various assays. We aim to enable research, like that performed by Jain et al. and others, in order to drive accelerate scientific discovery.
In addition to data capture and primary analysis, the Platform for Science can enable intelligent antibody research through R integration. The ability to seamlessly load, analyze, and store data within the platform provides scientists with endless visualization and analysis options. For example, Principal Component Analysis (PCA) is a powerful technique used to limit the redundancy in variables. By limiting experimental parameters, scientists can reduce experimental cost and arrive at a decision quickly. The Platform for Science also supports supervised and unsupervised machine learning that can predict outcomes from multivariate data sets. With these powerful techniques and many more, you will be able to transform your structured antibody data into predictive models and push research and science forward.
Antibody Profiling Applications for the Platform for Science
In support of research in the large molecule field, we have released 12 new applications to cover work in large molecule profiling. Leveraging the 12 assay types referenced in Jain et al. (2017), these applications can be used to record protocols, capture antibodies, platemaps, %I, fold change, thermal shift, and more. By combining the data from these assays within the LIMS, you will gain an overview of your antibody profile and can easily compare it to other antibodies in your database to make decisions quickly.
Antibody Profiling Applications:
These antibody profiling applications can be found in the Platform for Science Marketplace. Click on the app tiles to learn more.
Monoclonal Antibodies Approved by the EMA and FDA for Therapeutic Use (Status 2017), Animal Cell Technology Industrial Platform, 18 May 2017, www.actip.org/products/monoclonal-antibodies-approved-by-the-ema-and-fda-for-therapeutic-use/.
Jain, Tushar, et al. “Biophysical Properties of the Clinical-Stage Antibody Landscape.” Proceedings of the National Academy of Sciences, vol. 114, no. 5, 2017, pp. 944–949., doi:10.1073/pnas.1616408114.
Ecker, Dawn M, et al. “The Therapeutic Monoclonal Antibody Market.” MAbs, vol. 7, no. 1, 2015, pp. 9–14., doi:10.4161/19420862.2015.989042.
Booth, B. (2017, May 11). Human Antibody Discovery: Of Mice And Phage [Web log post]. Retrieved from https://lifescivc.com/2017/05/human-antibody-discovery-mice-phage/
Heather Adinolfi is the Biopharma Application Manager for the Platform for Science. She holds a Masters degree in Molecular and Cell Biology and 8+ years’ experience in genomics and cell biology laboratories. Most recently, Heather was a member of the Leads Discovery department at Bristol-Myers Squibb where she maintained and developed new cellular reagents for high throughput screening initiatives.