Closed Kcjohnson closed 5 years ago
Another consideration that has been discussed within the group is that it would be advantageous to create a flag to identify the higher quality set (among WGS and WXS) when both data types are available. In the event that the WXS and WGS samples for a particular subject appear of similar, preference for the sample included in the final analysis should be the WGS set as it presumably has a greater amount of information.
Going to close this for now. We addressed the purity issue in separately for different analyses so we did not set a hard threshold for a purity cut-off (e.g., purity > 0.5 for neutrality analyses). Eventually, we plan to incorporate another purity metric. We can revisit any filtering/blocklisting at that time.
During the 2018.12.13 teleconference Joe Costello (UCSF) asked whether we are block listing samples based on low sample purity. We had not previously block listed samples based on this criteria.
Investigating the sample with the lowest purity:
GLSS-MD-0003-R2-01-NB-01D-WGS
showed that it had a lower than expected mutation rate (0.01 coverage adj. mutation frequency) compared with itsR1
andR3
tumors that had purity values greater than 0.7.Most samples have greater than 50% tumor purity in this dataset so it is not expected that this filter will impact much of the cohort.
N.B. Other groups have previously run ABSOLUTE on the paired TCGA WGS/WXS samples. The correlation between TITAN and ABSOLUTE purities for these samples was Spearman's rho 0.60, P = 3.2E-10.![tcga-titan-absolute-cor-correctfalse](https://user-images.githubusercontent.com/6731211/50117908-30a57380-021c-11e9-806e-5dbd81b2f77c.png)