Abstract

An Advanced Statistical Description of Digital Quantification Methods and its Implications for the Design of Point-of-Care Devices

Manuel Loskyll, PhD student, Robert Bosch GmbH

The present work deals with an advanced statistical description of partition-based digital quantification methods like digital polymerase chain reaction (dPCR). Based on the combinatorics of the partitioning and detection process, we infer the methodologically inherent statistics and deviate a distribution for the evaluation of digital quantification experiments (DEDQuEx). The DEDQuEx allows not only giving the most probable result based on the observed number of positive partitions but also determining the statistical uncertainty of this result. We demonstrate that, particularly in the regime of low mean copy load per partition, dPCR systems provide a lower quantification uncertainty than previously expected. Furthermore, the DEDQuEx is combined with the statistics of subsampling in order to include its effects into the calculation and to elucidate its relevance. The results are applied to point-of-care systems with a certain sample volume and a limited area of detection to find guidelines how to optimally design such a dPCR system. It turns out that, especially in case of low target molecule concentration, the subsampling error is more prevalent than the uncertainty due to the partitioning. Consequently, dPCR systems should be capable of transferring the entire sample into the analyzed partitions to achieve maximum accuracy.