Completed on 18 Aug 2017 by Biswapriya B. Misra . Sourced from http://www.biorxiv.org/content/early/2017/08/17/171967.
Login to endorse this review.
Very useful tool. Looking forward to using it.
1. Would be useful if the Abstract points to the Webtool, a lot convenient to readers: https://metabolomics.cc.haw...
2. Of course, in metabolomics, missing values that exist in more than 20% of samples may be removed directly from the data, which is called “80% rule” (Bijlsma et al. 2006) vs. 50 % missing values for rejection of the variable is the case for Xia et al. MetaboAnalyst! Literature still flooded with 50% missing value cases. : (
3. "first data set included a total of 977 de-identified subjects and 75 metabolites": Source of this data? An accession number from MetaboLights/ Metabolomics Workbench would be useful. Which study is this one from ?
4. Same "Yan Ni et al. 2015)":An accession number from MetaboLights/ Metabolomics Workbench would be useful for readers/ reviewers to get the data.
5. Which platforms were used to generate the data sets in 3, 4 ?
6. Title reads, MS-data imputation, do you include GC-MS and CE-MS based data as well for comparison? If so, why not...would be far more robust and validating...
7. Unsure what is "medium sample size data set for both label-free and labeled data evaluation." in the serum metabolomics study.
8. Challenges for imputation in targeted data sets and untargeted data..?
9. The 2nd study cohort has diseased and healthy subjects: so imputation done across groups or done within each group? On contrary to the first study- whether such dichotomy existed?