Closed krigia closed 1 year ago
There are various factors that could affect the number of HVFs within and across samples, not limited to: QC measures, differing local tumor composition, or technical issues. As we do not have the bandwidth to provide specific analytical advice in scenarios like this, I can supply you with resources that could provide some insight:
@gcyuan @RubD @mattobny
Hello Giotto team,
We run HVFcalculate using method = "cov_groups" for a set of 12 tumor samples. Please see below an example of the command line:
_F1_normalized_standard_6000 <- calculateHVF(gobject = F1_normalized_standard_6000, method = "cov_groups", expression_values = c("normalized"), zscore_threshold = 1.65, show_plot = F, save_plot = T, save_param = list(save_folder = "hvf", save_name = "F1_normalized_standard_6000_hvf_covgroups"))
For the 12 FFPE samples from the same tumor type with similar QC results and a variable number of HVFs with a range 550-1000 (median ~800 HVFs), we observed a very low number of common HVF genes (range 1-13 common HVFs for different combination of samples). Particularly for one case with two tumor samples from the same individual, we only found 11 common HVF out of ~800 HVF genes. I understand HVFcalculate identifies the highly variable genes by definition for each sample, but I would expect to see some similarities regarding HVFs across all or subset of samples to some extent. Have you observed similar results in your analyzed cohorts?