mahmoodlab / PathomicFusion

Hi , Chen.

I have a question about the grade classification task.

According to the grade data.csv file, which includes age, gender, histologic grade, and subtype data

I understood that features you used to classify the grade included the subtype, correct?

Accordingly, the features for the grade classification task are: age, gender, subtype data, and the label is the histologic grade?

Thanks,
Omnia.

Hi @omniaalwazzan - grade (or any clinical metadata) is not used in the downstream tasks.

The clinical metadata is separated from the genomic features.

PathomicFusion/core/utils_data.py

Line 46 in 44b5513

    
           def getCleanGBMLGG(dataroot='./data/TCGA_GBMLGG/', ignore_missing_moltype=False, ignore_missing_histype=False, use_rnaseq=False, use_ag=False):

Thanks for answering @Richarizardd

Yes that makes sense!

I was quite confused as I used the exact function you referred (line 49) which joins the grade with genomics data; so I thought you were using them :)

Moreover, the Fig.3(A) shown the grade with the molecular subtype puzzled me a bit as if molecular subtype was accounted as one of the features!

Thanks,
Omnia.

Ah apologies for the confusion @omniaalwazzan. In Figure 3A, it is showing how the risks are distributed across the different histomolecular subtypes, from which you see that IDHwt-astrocytomas are prognosticated with higher risk scores.

Moreover, we see that the grade also correlates with the subtypes, e.g. - most of the IDHwt-astrocytomas that are predicted high risk are also Grade IV.

All in all, this plot is to help corroborate that the predicted risk scores corroborate with known biomarkers and current grading paradigms.

Thanks a lot for this clarification!

My confusion is resolved.

Is the molueculare subtype feature included in the Grade classification task?