Skip to content

Home

Jump to bottom
hnolCol edited this page Sep 26, 2018 · 1 revision

Welcome to the TrackQC wiki!

To setup TrackQC we highly recommend that one person defines all the settings and creates the models. Then other staff members can simply copy the unzipped folder of TrackQC, adjust the directories. Like this every researcher works with the exact same setting setup and models for outlier detection:

  1. Collect the search engine results in a folder of QC runs that you would classify as good QC runs
  2. Build thresholds and models instrument specific for the collected QC runs. Note – you can also specify a common model / threshold for multiple instruments of the same type. If the LC system differs, we would suggest to however to keep them separate.
  3. Select the directory in which you have saved all QC runs and click on Go to analyze the collect QC runs.
  4. Specify the instrument.
  5. On a normal desktop computer one QC run should take approximately 5 seconds.
  6. TrackQC will produce a pdf report for each QC run.
  7. Based on this set of QC runs the following things will be computed: a. For each specified metric (in the configuration tab) the 25 50 and 75 % quantiles will be calculated. The 20% and 75% quantiles are used to calculate the inter quantile range (IQR). This will be used to detect metrics that are out of range. As in a boxplot visualization an outlier is considered to be further away from the median than 1.5 x IQR (up or down). In every pdf report, the number of metrics that are out of the range will be indicated in the title. b. Added QC metrics will be plotted in the Overview plot. c. A trained model for outlier detection based on the Local Outlier Factor or Isolation Forest algorithm. Every newly imported QC run will be evaluated by this model. The metrics used for outlier detection can be specified by the user in the settings tab. The metrics values will be transformed to unit variance. Notably, if more than two metrics are selected (by default the MaxQuant naming: Peptide Sequences Identified and Score will be used) the metrics showing the highest variance will be used for visualization. Each pdf evaluation report for the specified instrument will contain from now on a plot that might look like this for an outlier.

Clone this wiki locally