Stephen Master (Presenter)
Weill Cornell Medicine
| Short Abstract The development of novel, highly multiplex assays has led to the possibility of new clinical assays with improved sensitivity / specificity, decreased sample size requirements, and reduced assay cost per analyte. While these clinical assays have potential diagnostic advantages, they have raised a new set of problems for traditional laboratory quality control (QC) paradigms. We will present four multiplex QC paradigms that address these challenges. Further, we will discuss the fundamental distinction between patterns (multiplex tests with associated bioinformatics classifiers that yield one of a small number of diagnostic outcomes) and panels (multiplex tests that are the equivalent of a series of uniplex tests). |
Long Abstract
The development of novel, highly multiplex genomic and proteomic assays has led to the possibility of new clinical assays with improved sensitivity / specificity, decreased sample size requirements, and reduced assay cost per analyte. While these clinical assays have potential diagnostic advantages, they have raised a new set of problems for traditional laboratory quality control (QC) paradigms. Current laboratory practice relies on a set of standardized quality control thresholds, such as Westgard rules, that ensure that changes in assay accuracy and/or precision are detected. However, multiplex testing creates a problem for these clinical paradigms that is analogous to the multiple hypothesis testing in research testing. Specifically, once a sufficient number of analytes are tested in a simultaneous multiplex assay, the false rejection rate for QC testing precludes the use of traditional metrics.
Although this statistical challenge initially appears to prevent the successful adaptation of clinical-grade QC, we will present four multiplex QC paradigms that address this issue. However, these paradigms must be specifically tailored to the type of assay. We will discuss the fundamental distinction between patterns (multiplex tests with associated bioinformatics classifiers that yield one of a small number of diagnostic outcomes) and panels (multiplex tests that are the equivalent of a series of uniplex tests). We will demonstrate that different QC strategies are appropriate for these distinct scenarios. Using Monte Carlo computational simulations, we will demonstrate an approach to the rational design of pattern-oriented QC using perturbation of the associated bioinformatics algorithm. For panel measurements, we will demonstrate the effects of repeat measurements on QC statistical power. Overall, we will demonstrate that even in the “worst-case” scenario where the quality of individual analytes in a multiplex assay are uncorrelated, we can develop efficient strategies that provide QC that is equivalent to current uniplex strategies used in the clinical laboratory.
References & Acknowledgements:
| Description | Y/N | Source |
| Grants | no | |
| Salary | no | |
| Board Member | no | |
| Stock | no | |
| Expenses | yes | MSACL, AACC |
IP Royalty: no
| Planning to mention or discuss specific products or technology of the company(ies) listed above: | no |