Premium
Peril beyond the winner's curse: A small sample size is the bane of biomarker discovery
Author(s) -
Thambisetty Madhav,
Casanova Ramon,
Varma Sudhir,
Legido Quigley Cristina
Publication year - 2017
Publication title -
alzheimer's and dementia
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 6.713
H-Index - 118
eISSN - 1552-5279
pISSN - 1552-5260
DOI - 10.1016/j.jalz.2017.01.003
Subject(s) - library science , psychology , medicine , computer science
We thank Mapstone et al. for their comments on our failure to replicate their index findings on plasma metabolite biomarkers of incident Alzheimer disease (AD) [1]. They raise three main concerns with our study [2] which they suggest precludes comparison with their earlier report. We respond to each of these in the following sections: 1. Plasma versus serum differences in metabolite profiles The most pertinent question in comparing methodological differences between the two studies is whether differences in the concentrations of the 10 metabolites between plasma and serum are large enough to explain the discrepant results. We believe that Table 4 in our study answers this question directly. It provides the actual concentrations of the 10 metabolites in plasma and serum as well as the correlations between the two matrices for each metabolite. Seven of the 10 metabolites show plasma versus serum correlation 0.86 when measured by an identical flow injection analysis mass spectrometry (BIOCRATES AbsoluteIDQ ) method as was used in both our study as well as by Mapstone et al. Moreover, even the two metabolites from the Mapstone panel, that is, C16:1-OH and PC aa C40:2 that were considered to have low measurement stability have similar concentrations in plasma and serum [3]. To the best of our knowledge, these data from Yu et al. are the only head-to-head comparison between plasma and serum tested on the same BIOCRATES AbsoluteIDQ metabolomics platform used both by Mapstone et al. as well as in our recent publication. The publications cited by Mapstone et al. in support of their contention that plasma and serum provide varying metabolite measurements are from studies using very different methodologies, that is, gas chromatography/time-of-flight mass spectrometry and ultrahigh performance liquid chromatography mass spectrometry [4,5]. Both these methods are applied in untargeted and semiquantitative metabolite profiling, unlike the BIOCRATES AbsoluteIDQ platform that is