Multiplex immunoassays (MIas) are moderate- to high-throughput platforms for simultaneous quantitation of a panel of analytes, and have gained popularity as hypothesis generating tools for targeted biomarker identification. For diagnostic or screening purposes, optimized single-analyte immunoassays yield concentration estimates from a calibration curve's linear range, with standards bracketing 50% effective concentration in log-increments. But as tools for biomarker discovery, MIas are not always rigorously validated in the target population, and often little is known about an analyte's expected concentration in samples derived from that population. As a consequence, MIa data are often plagued by high proportions of concentrations flagged either as 'out-of-range'- samples for which the observed response falls below (above) the lower (upper) asymptote of a non-linear calibration curve - or as extrapolated beyond the smallest or largest standard. Small out-of-range or extrapolated concentrations are left-censored because the true concentration is known only to be less than the minimum standard;large out-of-range or extrapolated concentrations are considered right-censored. In addition, analytes are targeted according to a putative biologic pathway and their concentrations are likely to be correlated, yet traditional analysis techniques ignore this multivariate structure. Yet, as the number of analytes assessed in a given experiment is small (tens of markers), accounting for the multivariate nature is feasible and would be expected to improve precision by drawing strength across the markers. Furthermore, it is reasonable to expect that statistical methods harnessing important co-variation among markers are more likely to uncover important biological structure than those assessing one marker at a time.
The aims of this proposal address the major data analysis barriers to the utility of MIas in oral health biomarker discovery research, specifically: 1) large proportions of censored observations;2) differences in the precision (as quantified by %CV) with which concentrations are estimated;and 3) the data's inherent multivariate structure.
The aims of this proposal address the major data analysis barriers to the utility of multiplex immunoassays in oral health biomarker discovery research, specifically: 1) large proportions of censored observations;2) differences in the relative precision with which concentrations are estimated;and 3) the data's inherent multivariate structure. Achievement of the proposed aims will provide oral health researchers and their statistical collaborators the methods needed to correctly analyze censored multiplex immunoassay data and enhance the potential for discovery of biologically meaningful structure, thereby increasing the utility of this platform in targeted biomarker discovery research.