Since biological therapeutics are derived from living organisms, their manufacture and validation presents difficulties not encountered during traditional small molecule drug development. Despite the complexities inherent in developing biologics, there are a growing number of licensed biological products on the market.
IBC’s “Bioassay and Development” meeting held earlier this month in San Francisco addressed the challenges facing researchers in academia and industry. Bioassays are critical for the development of biologics, and many companies now find themselves in a regulatory holding pattern because of poorly developed and inadequately validated potency assays.
To correct this problem, many biotech and pharma firms have earmarked a significant portion of their R&D budgets for developing high-performance bioassays that measure potency accurately and achieve product acceptance quickly with minimal cost.
Many factors affect bioassay performance, explained David Lansky, Ph.D., president of Precision Bioassay. At the meeting, Dr. Lansky spoke about how mismatches between bioassays and models result in imprecision. “Pseudoreplication during sample generation is commonplace in assay design and leads to misleading estimates of precision.”
Precision Bioassays’ Xymp™ bioassay system is a solution to the problem, he said. “It reduces cost, labor, time-to-market, and developmental and regulatory risk.” The three-module software system has an optional component that utilizes robotics to facilitate good plate layouts using statistical designs that implement structured randomization.
“Blocking is a hugely powerful tool, and including the 96-well plate as a block is a useful idea that escapes most bioassay design models,” Dr. Lansky said. He explained that “randomization by the Xymp robotic system allows for the use of more complex software models to analyze the data.”
Xymp’s second module provides for sophisticated outlier detection, which is important as removal of outliers before re-scaling the data is a poor strategy, according to Dr. Lansky. “Re-scaling the data usually removes 80 percent of the problems and restores at least near symmetry if not near normality. Researchers need to look at the sources of the variance.”
A common approach in bioassay analysis is to examine small groups of replicates (or pseudoreplicates) for outliers. A much better approach uses all of the data in the assay together, but with milder assumptions about the shape of the dose-response curve (i.e., a smooth curve) than used in the models fit to assess similarity and estimate potency. The use of reproducible randomized experimental designs with more samples allows more complex statistical methods to identify outliers.
The third Xymp module implements equivalence testing for similarity via mixed-effects models and delivers graphical summaries and potency estimates, Dr. Lansky explained, adding that there are often important components of variance associated with sample assignment to rows or dilution assignment to columns that mixed-effects modeling addresses.