A Digital PCR Data Analysis Workflow: From Sample to Effective Result

Digital polymerase chain reaction (dPCR) is a relatively novel biotechnological method that allows for the quantification of DNA/RNA molecules. Due to its reported improvements (increased accuracy, precision, resistance to unwanted effects, etc.) over its predecessor quantitative PCR, dPCR is being adopted by an increasing number of life sciences researchers. The path from raw data (generated from a biological specimen) to a final and interpretable result is, however, littered with obstacles. There is an increasing body of literature discussing several issues in the dPCR data analysis workflow. Here, we present some of the major issues. Based on our own work and that of others, we discuss the potential and drawbacks of several statistical methods that have been proposed to deal with these oft-ignored problems: initial thresholding ((robust) clustering, extreme value theory, nearest neighbours), combining technical replicates and/or multiple biological samples into a meaningful estimate (generalized linear mixed models) and final biological results (comparison of patients with a healthy reference group). We demonstrate the workflow on a cancer patient dataset, consisting of samples obtained from a group of healthy reference individuals and samples obtained from cancer patients. The aim is to study the discriminatory power of copy number alteration as measured by dPCR.

http://ift.tt/2qgQGgU