Open AccessReproducibility challenges for biomarker detection with uncertain but informative experimental data
Author(s) -
Wei Vivian Zhuang,
Luísa Camacho,
Camila S. Silva,
Huasheng Hong
Publication year - 2020
Publication title -
biomarkers in medicine
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.652
H-Index - 44
eISSN - 1752-0371
pISSN - 1752-0363
DOI - 10.2217/bmm-2019-0599
Subject(s) - biomarker , reproducibility , medicine , inference , biomarker discovery , computational biology , data set , data mining , set (abstract data type) , microrna , bioinformatics , computer science , statistics , artificial intelligence , gene , biology , mathematics , proteomics , biochemistry , programming language
Recent studies have revealed that circulating microRNAs are promising biomarkers for detecting toxicity or disease. Quantitative real-time polymerase chain reaction (qPCR) is often used to measure the levels of microRNAs. Besides complete and certain data, investigators inevitably have observed technically incomplete or uncertain qPCR data. Investigators usually set incomplete observations equal to the maximum quality number of qPCR cycles, apply the complete-observation method, or choose not to analyze targets with incomplete observations. Using biostatistical knowledge and published studies, we show that three commonly applied methods tend to cause biased inference and decrease reproducibility in biomarker detection. More efforts are needed to address the challenges to identify and detect reliable, novel circulating biomarkers in liquid biopsies.