3. Comparing Tumor vs. Normal Samples

1. To compare Tumor samples vs. Normal samples, you first need to join the Sample information to the Integer Segment Means spreadsheet. To do this open MIP_Sample_Table Dataset - Sheet 1 and select File > Join or Merge Spreadsheets. Select Segment Covariates Every Column and click OK. This will pop up the Join or Merge Spreadsheets window with various parameters to ensure the two spreadsheets are properly joined.

   In most cases we recommend joining on row labels as both spreadsheets should contain the same unique identifiers as row labels. However, this may not always be the case and therefore you may need to use the sorting option to join them. To do this, select Use a custom sort order under Matching Criteria and then click Define to define the sort (Figure 8).

   

   Figure 8. Custom sort order

   The two spreadsheets should already be in the same order though it would be good to double check. To make subsequent analyses easier, choose Use left spreadsheet labels under Row Labels for Joined Spreadsheet, and click OK. Back in the Join or Merge Spreadsheet window, enter Sample Info + Segmentation Covariates Every Column as the New dataset name. It should look like Figure 9. Click OK to finish the join.

   

   Figure 9. Join or Merge Spreadsheet window with a custom sort order selected

2. Now create a Heat Map of the data. Open the Sample Info + Segmentation Covariates Every Column - Sheet 1 spreadsheet and select Plot > Heat Map. This will generate a Heat Map of all the samples with the default three-color scheme (Figure 11).

   

   Figure 11. Default heat map of all samples

3. From the Graph Control Interface (upper left pane) click on the graph item Sample Info + Segmentation Covariates Every Column this will bring up the graph item controls. Click on the Group tab and select Select variable.... Choose Sample Type from the list and then click OK.

   You will now have all “Normal” samples in the top group as indicated by the blue bar on the left, and all “Tumor” samples in the bottom group as indicated by the green bar on the left. (Figure 12) The tumor group can be moved to the top by clicking and dragging the group in the Group box and dropping it on top of the “Normal” group.

   

   Figure 12. Heat map grouped by sample type

   To make it easier to refer to this heat map, right-click on the Sample Info + Segmentation Covariates Every Column graph item in the Graph Control Interface and select Rename. Change the name of this node to Normal vs Tumor Samples.

4. Now change the color scheme of the heat map from the default to CNV defaults. Click on the Normal vs Tumor Samples graph item in the Graph Control Interface (upper-left pane). Click the Color tab. Click Manual and then click Set CNV Defaults.

   The CNV defaults actually represent common copy number boundaries based on log ratio values, which are centered about zero for most array platforms. The MIP array is a little different where the values are centered around two, where any value below 1.5 is considered a copy number loss, and any number above 2.5 is considered a copy number gain.

   So change the color scheme to: Red = 1.5, White = 1.6, White = 2.4, Green = 2.5. You can change the numbers by first selecting the Manual radio button, then clicking and hovering above each one.

   You will now see 3 colors in the heat map: Red = 1+ copy loss, White = Neutral, Green = 1+ copy gain. You can also begin to see streaks of copy number variation in the Tumors group and not in Normals (Figure 13).

   

   Figure 13. Heat map set to MIP copy number scheme

   You can zoom into areas by left-clicking and dragging on the graph or X and Y axes, by double-clicking a chromosome number in the Full Domain view or a cytoband/gene in the Annotation Tracks (bottom), by selecting a chromosome from the drop-down menu in the tool bar, by manually entering a chr/position string (e.g. chr1:10000000 - chr1:100000004) or a search string (e.g. 1q21 or APOE) in the Genome Location Bar in the tool bar.

   Figure 14 is zoomed into chromosome 21 where large green streaks (gains) appear across a majority of the Tumor samples and not in the Normals.

   

   Figure 14. Zoomed view of chromosome 21 shows copy number changes in tumor samples that do not appear in normal samples

5. There’s also a nice streak of copy number loss on 7p14.1 that appears to show a loss of the TARP gene (transcripts NM_001003806 and NM_001003799) and perhaps the adjacent gene STARD3NL (transcript NM_032016). You can investigate this gene further by clicking the gene in the Annotations pane and selecting a contextual web link in the Data Console.

   

   Figure 15. Region on 7p14.1 that shows copy number loss in most tumor samples but not in normal samples

6. The ability to change color schemes on the fly is pretty powerful with the MIP array where you can have numerous multi-copy gains. To differentiate between multi-copy gains you can add additional colors (one color for each additional copy). If you only wanted to investigate multi-copy gains, you could, for example, delete the Red = 0 item and change White to any number greater than 4 and Blue to 10. This allows you to focus on the more extreme copy number gains.