Closed andrewelamb closed 5 years ago
ok, getting to this
OK, looks like one of the subtypes of AML (AML.1) doesn't have a member with status ==0 ... might be causing it... here's a table with counts of each status for each subtype. Can probably fix this early tomorrow.
> # A tibble: 8 x 3
> variable Num1 Num0
> <fct> <int> <int>
> 1 AML. 6 2
> 2 AML.1 12 0
> 3 AML.2 14 5
> 4 AML.3 5 13
> 5 AML.4 21 15
> 6 AML.5 20 13
> 7 AML.6 19 9
> 8 AML.7 14 6
>
I have narrowed it down to the call to create_kmplot... in particular, error happens when risk.table = T.
There is no error by removing the group_colors. For some reason, with AML, there's 8 subtypes but 14 colors in the group_colors here. So, a quick fix is to remove it.
survminer::ggsurvplot( fit, data = df, conf.int = F, risk.table = T, title = "title", palette = group_colors )
The second error, in the heatmap, comes from the ci_mat matrix, which is all NaNs.
Several of the features (columns) in LAML, are all NAs ...
FIXED the color bug!
Problem was the survival fit appended the subtype names with the column name, "variable=", and then the plot function needed palette to be a named vector --- matching the subtype names --- like 'variable=LAML.1'
The second bug in the heatmap ... I think just comes from some TCGA studies that have missing info in the feature matrix.... can maybe just catch it and display a message.
Certain TCGA subtypes such as 'AML' cause errors like:
'names' attribute [8] must be the same length as the vector [4]
When the Clinical Outcomes module attempts to create a kmplot. These seem to be smaller groups and may need to handled as an error.