Application of Design of Experiments using JMP to

Application of Design of Experiments using JMP to Develop a Universal Quantitative Method for Preclinical Pharmacokinetic Support Alok Rathi, Carina Carter, Qiang Qu, Jim Mc. Nally, Theresa Goletz. EMD Serono Research and Development Institute, Inc. , Billerica, MA Introduction Pharmacokinetic (PK) assays that measure the drug concentrations in a sample, are essential to the non-clinical safety data package. PK assays for large molecules require the generation of reagents specific to the biotherapeutic. A universal quantitative assay that recognizes an epitope of a given framework common to various biotherapeutics, for example the common human Ig. G framework, would improve efficiency by reducing reagent generation costs, minimizing resource utilization and shortening timelines. Multiple commercially available anti-human Ig. G antibodies, which specifically bind to any molecule containing the human Ig. G framework epitope, were screened for their binding affinity to the reference human Ig. G. Instead of using a one factor at a time approach, a more efficient design of experiment (DOE) approach was utilized using JMP to optimize assay parameters that achieve targeted responses. Multitude of factors were tested simultaneously, such as assay buffers, range of minimal required dilutions and concentrations for capture and detection reagents. This significantly reduced the number of experiments that would be needed to test each factor separately, thereby considerably shortening the assay development time. The DOE approach using JMP allowed for a rapid development and qualification of a universal assay in multiple species. Objective • • • A ligand binding enzyme linked immunosorbent assay (ELISA) experimental design usually requires the need to identify the capture and detector antibodies, and then optimize the conditions of these critical reagents, so that they interact with the drug molecule, to achieve the desired results. Image 1: A standard sandwich ELISA method The optimization process includes testing multiple parameters that determine the assay outcome and is mostly based on a trail and error process using previous knowledge and experience. The design of experiments (DOE) approach provides a systemic method to determine the relationship between multiple factors affecting the experiment and their effect on the response for that experiment. The DOE approach was used to evaluate which assay inputs (i. e. buffers, concentrations, dilutions etc. ) had a significant impact on the process output, and what the target level of those inputs should be to achieve the desired output. Testing multiple factors simultaneously increases efficiency and decreases the time and resources required to complete assay development. Screening Reagents From a panel of commercially available anti-human Ig. G antibodies, a pair was chosen, based on the binding affinities of the antibodies to the Ig. G molecule, determined by Octet, and which gave higher sensitivity and a broader dynamic range by testing in a checkerboard ELISA format on the Meso Scale Discovery (MSD) platform using default conditions. • Image 2: Pairing experiment on the Octet • • Image 3: Checkerboard ELISA run on MSD platform result DOE – Full Factorial Design approach was utilized to optimize assay development decisions on multiple factors simultaneously. • Ø Ø Ø Factors tested: Capture concentration Detector concentration Minimal Required Dilution (MRD) Ø Assay buffers • Ø Ø Ø • Image 5: Table with each run information Design of Experiment setup (DOE) in JMP • A plate map was carefully designed by grouping different factors based on the run information table for better execution. Only three samples were needed for each run : HQC, LQC and a Blank sample. The concentration of the HQC and LQC samples were chosen based on the targeted sensitivity and the range of quantification. The data from the plate was then used to have the statistical analysis using JMP. • Image 4: Full Factorial DOE setump in JMP 13. 0. 0 Responses High Quality Control (HQC) / Low Quality Control (LQC) / Blank samples Maximizing the signal / noise ratio of the responses would result in better sensitivity, lower background a broader dynamic range of quantification. DOE Execution • Table with run information was generated after DOE design was setup. • 2 responses for 4 factors 2 x 2 x 2 x 2 = 16 total runs. Image 6: Plate map designed based on the run information table Results • • • Ø Ø After data input for each run in the run information table, the data was analyzed by selecting the model specifications, and the prediction profiler was obtained. The prediction profiler finds the factor settings that maximize the predicted yield. The change in certain factors, as can be analyzed by movement of the red dotted line in the profiler, would significantly affect the assay responses. Desirability is maximized for each response, and the most balanced desirability is selected based on analytical understanding of the assay. The conditions that were determined by performing the DOE in a single run were: Coating Concentration of 5 µg/m. L. MRD of 1: 25. Detector Concentration of 0. 5 µg/m. L. BGG assay buffer. Image 7: Prediction Profiler Conclusions • The DOE approach allowed for the rapid development of the universal assay with a broad dynamic range, which can be applied to several drug molecules containing the human Ig. G framework in multiple non-clinical species. • Multiple factors were tested simultaneously using DOE, which allowed for determining the optimum conditions for the experiment in a short period of time, thereby significantly helping to shorten development timelines and minimizing resource utilization.

Image 1: A standard sandwich ELISA method 2

Image 2: Pairing experiment on the Octet to determine the antibodies with higher binding affinity to the human Ig. G • Binding affinity of the anti-human Ig. G antibodies to the human Ig. G were determined using octet, and the high binding pairs were selected to test on a plate based MSD platformat. 3

Image 3: Checkerboard ELISA run on MSD platform to determine antibody pairs Factors being considered to determining suitable capture and detector: • High S/N ratio • Low background signal 4 Title of Presentation | DD. MM. YYYY

Image 4: Full Factorial DOE setup in JMP 13. 0. 0 A full factorial design was setup to determine the optimum conditions for the assay. ØThe factors being tested were: • Coating concentration (Continuous – concentrations tested from 2µg/m. L to 5µg/m. L) • Detector concentration (Continuous – concentrations tested from 0. 5µg/m. L to 1µg/m. L) • MRD (Continuous – dilutions tested from 1: 10 to 1: 40) • Assay buffers (Categorical – factors tested in 2 different assay buffers – buffer containing Bovine Gamma Globulin (BGG) and commercial Superblock buffer) ØThe response desired were: • Maximum signal/noise ratio between High Quality Control (QC) / Low QC and Low QC / Blank samples • Maximizing the response would result in better assay sensitivity and would give a lower background a broad range of quantification 5 Title of Presentation | DD. MM. YYYY

Image 5: Table with run information based on the design setup in JMP 13. 0. 0 • The full factorial design setup required 16 runs to execute the DOE. • Various combinations for all the factors were to be tested, for JMP to be able to provide a model for the range of values of the factors, to achieve the desired responses. • The runs can be grouped together for any one factor for convenience in experiment execution. 6 Title of Presentation | DD. MM. YYYY

Image 6: Plate map designe based on the run information table in JMP • The plate map being carefully designed by grouping the capture concentrations together in the run information table in JMP design setup. • Only three samples were needed for each run to calculate the required responses set in JMP. • The concentrations of the high and the low quality control samples (HQC and LQC) were chosen based on the targeted sensitivity and range of quantification desired. 7

Image 7: Prediction Profiler – to determine the factor settings that maximize the predicted yield. • The profiler finds the factor settings that maximize the predicted yield. • The change in certain factors, as can be analyzed by movement of the red dotted line in the profiler, would significantly affect the assay responses. • Desirability is maximized for each response, and the most balanced desirability is selected based on analytical understanding of the assay. • The conditions that were determined by performing the DOE in a single run were: • Coating concentration is 5 µg/m. L • MRD is 1: 25 • Detector concentration is 0. 5 µg/m. L • Assay buffer is buffer containing Bovine Gamma Globulin (BGG) 8