Question 1

Why does cohort design matter more than instrument choice?

Accepted Answer

Age, sex, body composition, diet, medication and fasting status influence the metabolome strongly, often more than the disease under study. If cases and controls differ systematically in any of these, the analysis measures that difference instead. Match on the major factors, record the rest, and adjust for them explicitly.

Question 2

How are collection artefacts avoided?

Accepted Answer

Tube type, time before centrifugation, storage temperature and number of freeze-thaw cycles each leave measurable signatures. Collect all groups under one protocol, ideally in the same clinics, and record handling details per sample. A study where control samples came from a different biobank cannot separate biology from processing.

Question 3

Should the study be untargeted, targeted, or both?

Accepted Answer

Untargeted profiling finds unexpected signals but gives relative values and many unidentified features. Targeted panels quantify known molecules accurately with calibration and internal standards. The usual structure runs untargeted discovery, then builds a targeted assay for the promising subset and applies it to the replication cohort.

Question 4

What quality control belongs in a long analytical run?

Accepted Answer

Randomise sample order across groups, inject pooled quality control samples at regular intervals, include extraction blanks, and monitor internal standard response. Apply drift correction based on the pooled samples and report how many features met the stated reproducibility criteria, since unstable features should not enter the analysis.

Question 5

How is overfitting avoided in multivariate models?

Accepted Answer

With thousands of features and few samples, a model can separate any two groups perfectly by chance. Keep a held-out set untouched during development, use nested cross-validation for tuning, and run permutation testing to establish a null distribution. Report performance on data never used in fitting.

Question 6

What identification confidence should be reported?

Accepted Answer

An accurate mass gives a formula at best, and isomers are common in metabolism. Matching retention time and fragmentation against an authentic standard on the same system is the strongest evidence. State the confidence level reached for each reported marker rather than presenting tentative assignments as identified molecules.

Question 7

What does replication actually require?

Accepted Answer

An independent cohort, preferably recruited elsewhere with different handling, analysed by a quantitative targeted method with prespecified thresholds. Report discrimination and calibration rather than significance alone, and compare against existing clinical markers, since a marker that adds nothing beyond routine measurements has limited practical value.

Biotechnology

Biotechnology

Research Frontiers in Metabolic Disease Biomarker Discovery