QA Checklist for Cancer Browser: Difference between revisions

From genomewiki
Jump to navigationJump to search
m (add category tag)
Line 4: Line 4:
== QA checklist for tracks on the cancer browser ==
== QA checklist for tracks on the cancer browser ==


* Tracks can have both Chromosomes and Genesets as display options, check both
** Occasionally tracks will work in one but not the other, but give no warning of this (or only a little tiny on on their details pages)
*  Speed. Does it take a long time for a dataset to load? Indicating there is no down-sampled data table, or data table is not indexed.  
*  Speed. Does it take a long time for a dataset to load? Indicating there is no down-sampled data table, or data table is not indexed.  
 
*  Is there data on every chromosome?  Any missing sections should be accounted for.  
*  Is there data on every chromosome?  Any missing sections should be accounted for. Sometimes X/Y. 
* Do the colors look okay? Are there any visual issues with the track?   
**is the overall genome-wide view color looks too dim or too bright?
*  Does the clickthrough to the Genome Browser worK?
**is the geneset view color looks too dim or too bright?
*  Test mostly using heatmap, but remember to check Boxplot and Proportions modes as well
**CNV data track : in the whole genome view, is the data looks like copy number data?
*  Clinical data: Features and Feature Settings:   
***If there is a gender clinical feature, does the genomic data on X chromosome matches the gender variable?
**How many features are on by default? Too many will cause feature names not to display
**Gene expression data: under whole genome view, is there any batch effect? typical gene expression data track does not show obvious pattern under whole genome view (there are exception such as if we normalized tumor againstnormal samples).
**Does sorting features work? Does sorting features properly sort the heatmap?
*  Zoom in and out
**Use features to sanity-check data. Does gender match with X/Y, that sort of thing
*  Drill down to base level.  Is the transition between the views sensible?
**View Feature info by using the blue button next to the Feature graph. Here, check for:
*  Do the graphics make sense?   
***Feature values: do they match what gets graphed? Do they make sense? Sometimes non-numerical values get made into a numerical scale, for example.
*  Does the clickthrough to the Genome Browser work?
***Do the Feature short labels match up with their names?
**go to the correct spot?
***Are there repeated features, or features with meaningless info (ie gender in an Ovarian tumor dataset)
**correspond with its alignment track in the genome browser?
*** Stats function: Run one or two  statistical tests. Keep in mind that you have to designate subgroups in order for the function to work
**check with featureBits or Table Browser
*  Test mostly using heatmap, but don't forget to toggle over to summary view too.
*  Clinical data.  Fewer than 10 default features. 
* Do clinical sorts work? With microarray data?  Does genomic data sort work?
* Do the clinical features make sense?  And iss the legend/key sensible?
**I think the clarity of these labels is really important.  
**Also, in the subgrouping control, after you click on a feature, the feature categories/values will show up, do they make sense?
**Sometimes, they are just left as numerical (float) numbers (showing as a slider in the UI) , it makes sense for patient age, gi50 sort of features, but not for features like TP53 mutation status kind of features.
* Stats function: Run one or two  statistical tests compare to subgroups under both whole genome and gene set view.
* Data track documentation:  After we commit Chris's help page implementation, is there a help page, and has the date source, pre-processing been documented?
* Data track documentation:  After we commit Chris's help page implementation, is there a help page, and has the date source, pre-processing been documented?
 
**I have no idea what this means
[[Category:Browser QA]]
[[Category:Browser QA]]

Revision as of 17:14, 26 July 2010

This is a rough draft for now


QA checklist for tracks on the cancer browser

  • Tracks can have both Chromosomes and Genesets as display options, check both
    • Occasionally tracks will work in one but not the other, but give no warning of this (or only a little tiny on on their details pages)
  • Speed. Does it take a long time for a dataset to load? Indicating there is no down-sampled data table, or data table is not indexed.
  • Is there data on every chromosome? Any missing sections should be accounted for.
  • Do the colors look okay? Are there any visual issues with the track?
  • Does the clickthrough to the Genome Browser worK?
  • Test mostly using heatmap, but remember to check Boxplot and Proportions modes as well
  • Clinical data: Features and Feature Settings:
    • How many features are on by default? Too many will cause feature names not to display
    • Does sorting features work? Does sorting features properly sort the heatmap?
    • Use features to sanity-check data. Does gender match with X/Y, that sort of thing
    • View Feature info by using the blue button next to the Feature graph. Here, check for:
      • Feature values: do they match what gets graphed? Do they make sense? Sometimes non-numerical values get made into a numerical scale, for example.
      • Do the Feature short labels match up with their names?
      • Are there repeated features, or features with meaningless info (ie gender in an Ovarian tumor dataset)
      • Stats function: Run one or two statistical tests. Keep in mind that you have to designate subgroups in order for the function to work
  • Data track documentation: After we commit Chris's help page implementation, is there a help page, and has the date source, pre-processing been documented?
    • I have no idea what this means
Retrieved from "https://genomewiki.ucsc.edu/index.php?title=QA_Checklist_for_Cancer_Browser&oldid=10992"