Sources of Error and Its Control in Studies on the Diagnostic Accuracy of "-omics" Technologies
Overview
Authors
Affiliations
Analyses of errors in diagnostic studies have led to improvements in the methodological quality of traditional laboratory research. However, since features of genomics and proteomics research ("-omics") differ from those of traditional research, sources of error are also likely to be distinct. We examine the main sources of error that are particularly relevant to "-omics"-based diagnostic techniques through the analysis of primary research papers which address these potential errors, their solutions, and the resulting spurious effect on diagnostic accuracy prediction. The main sources of error described in "-omics"-based research are mainly associated with chance: overfitting and inadequate sample size; variation: preanalytical variation (specimen collection and management), analytical variation (test procedures and reproducibility) and biological variation. We conclude that "-omics"-based research is prone to several errors. We have characterized them and shown the range of available solutions. This is a key step in the application of genomic discoveries to clinical and public health practice.
RIFS: a randomly restarted incremental feature selection algorithm.
Ye Y, Zhang R, Zheng W, Liu S, Zhou F Sci Rep. 2017; 7(1):13013.
PMID: 29026108 PMC: 5638869. DOI: 10.1038/s41598-017-13259-6.
Parker L, Saez N, Lumbreras B, Porta M, Hernandez-Aguado I PLoS One. 2010; 5(7):e11419.
PMID: 20625481 PMC: 2896422. DOI: 10.1371/journal.pone.0011419.
Towards proteome standards: the use of absolute quantitation in high-throughput biomarker discovery.
Chao T, Hansmeier N, Halden R J Proteomics. 2010; 73(8):1641-6.
PMID: 20399287 PMC: 2885480. DOI: 10.1016/j.jprot.2010.04.004.
Vineis P, Khan A, Vlaanderen J, Vermeulen R Environ Health. 2009; 8:54.
PMID: 19948053 PMC: 2793242. DOI: 10.1186/1476-069X-8-54.