Hierarchical Bayesian modeling identifies key considerations in the development of quantitative loop-mediated isothermal amplification assays

Abstract

Motivation:Loop-mediated isothermal amplification (LAMP) is a rapidly growing, fast, and cost-effective technique for detection of DNA/RNA in point-of-care biomedical applications. However, it remains unclear what factors affect LAMP's quantitative resolution, and experimental optimization of primers presents a major bottleneck in assay design. A lack of model-based frameworks to characterize LAMP data and address these questions presents an unmet need for LAMP assay development.Results:We present hierarchical Bayesian models of LAMP amplification based on Gompertz functions, and use these models to infer the effect of RNA variation and other factors on LAMP amplification curves derived from 80 blood samples of patients with suspected acute infection. Our analysis uncovers associations between LAMP assay resolution and characteristics such as primer sequence composition and thermodynamic properties. In addition to correlations between RNA input abundance and time shift of the the LAMP amplification curve, we also detect RNA-dependent assocations with amplification rate. We further investigate associations between primer/target properties and quantitative performance of the assay by generating a set of synthetic RNA samples with systematically varied primer sequences and applying our framework. We find evidence that the associations observed are driven by across-target rather than within-target variation, an important observation for study design. Our findings represent important first steps towards guided development of quantitative LAMP assays.Availability and Implementation:Analysis and modeling code is available upon reasonable request.