Better bioassays via designs for robots analyses with improved model selection and similarity bounds that limit potency bias

项目总结

Project Summary Biological assays (bioassays) provide critical measurements of the activity of complex biological drug products (proteins, vaccines, cell therapy, and gene therapy products) particularly during product development and production quality control. These assays use living systems for measurement and have special designs and analyses to accommodate the large variation and confounding factors involved. Common lab bioassays impose complex statistical structures that, if not designed and analyzed properly, lead to poor precision for estimates of potency and limited capabilities to monitor assays for developing problems. These problems contribute to lot release failures for manufacturers and slow down or block the development of biopharmaceutical products for consumers. Precision Bioassay provides statistical services for bioassays, helping organizations that develop, use, perform, validate, monitor, and regulate bioassays use modern methods to get better information and performance, ultimately improving the quality and accelerating the discovery of biopharmaceutical drug products. The proposed research will lead to three new tools that will improve bioassay design, analyses and monitoring, each will become a new product (a combination of software and consulting support). New bioassay designs that exploit the capabilities of modern robots will, in combination with good analyses, separate important sources of variation that are confounded with current designs and analyses; this improved information will accelerate bioassay development and enhance monitoring of bioassays. Improved methods for selection of which sources of variation to include in bioassay analysis models (equivalence-based random effects model selection) will improve estimation of important sources of variation that are often ignored with current analysis methods; this will substantially improve bioassay monitoring. Limiting bias of potency is essential to ensuring that a bioassay is fit for its intended use: a new approach to setting bounds for equivalence tests of similarity of test samples to standards will limit bias of potency associated with allowed non-similarity to pre-specified levels, ensure that most samples that are similar will pass similarity, and estimate the assay size required to achieve this performance target. In Phase II we will extend the proof-of-concept results from this research using prototype software able to address practical cases. The products from this research will become additional packages in our existing software and consulting services that will accelerate the development of biopharmaceutical products.