Skip to content

Data Review

An overview of data quality and of what the filters removed. It has four views: Summary, Feature plot, Tree map and CV plot.

Summary

The Data Review summary
The Data Review summary for the demo dataset.
  • Dataset: features in the peak table, features that passed every filter, and the numbers of injections, samples and biological groups used.
  • Data quality: three scores from the CV plot (below).
  • Filtering: how many features failed each filter. A feature can fail more than one, so the rows add up to more than the total.
  • Plot Feature Sets: how many features each set matched.
  • Analysis info: the run's parameters, the same text as analysisinfo.txt in Downloads.

Feature plot

Every feature in the peak table at its m/z and retention time, coloured by the filter that removed it. Features that passed every filter are drawn in black as High Quality.

The feature plot, coloured by filter outcome
Every feature in the peak table, coloured by the filter that removed it. Black features passed every filter.

Use it to see where in the chromatogram each filter acts. A filter that removes most features in one narrow region is worth a look in the raw data.

Tree map

The share of features removed by each filter, as areas.

The filtering tree map
Share of features removed by each filter.

CV plot

The cumulative percentage of features (y) against their coefficient of variation across technical replicates (x), drawn for both the median CV (blue) and the mean CV (red). The crosshair marks the CV threshold and the fraction of features under it.

The CV distribution plot
Cumulative distribution of median CV (blue) and mean CV (red). The crosshair marks the CV threshold.

How to read it:

  • A curve that rises steeply near zero means most features are reproducible.
  • A wide gap between the mean and median curves means the distribution is skewed: a few samples reproduce much worse than the rest.
  • A shallow curve, with few features under 0.2 to 0.3, often means alignment problems.
  • A sharp step partway up suggests inconsistent detection between replicate sets.

The three scores below the plot summarize it:

Score Meaning
Reproducibility Area under the curve, relative to the CV expected from counting noise alone for this number of replicates. Higher is better.
Skewness The gap between the mean and median curves. Lower is better.
Overall Reproducibility × (1 − Skewness), as a percentage.