Issue about Metabric dataset

[heyhola]

Hello Rahul,
Thanks for the great work ! I tried to use metabric dataset, but I have applied for a long time without passing. I see that you have this dataset here, but it is not marked. Could you tell me what x1-x9 in metabric.csv stands for?
Best wishes!
Siqi

[sourcesync]

@heyhola I had the same question. Look at the paper and especially this section:

METABRIC: The Molecular Taxonomy of Breast Cancer International Consortium
(METABRIC) is a clinical dataset which consists of gene expressions used to determine different subgroups of breast cancer. We consider the data for 1,904 patients
with each patient having 9 covariates - 4 gene indicators (MKI67, EGFR, PGR, and
ERBB2) and 5 clinical features (hormone treatment indicator, radiotherapy indicator,
chemotherapy indicator, ER-positive indicator, age at diagnosis). Furthermore, out
of the total 1,904 patients, 801 (42.06%) are right-censored, and the rest are deceased
(event). We obtained the DAG as depicted in Fig. 3 using a modified DAG-GNN
algorithm.

[rahulk207]

Hi folks, and sorry @heyhola for missing your comment. For METABRIC, please assume x1-x9 to be in the same order as mentioned in the above paper snippet shared by @sourcesync. However for GBSG, this isn't the case and the order is different than the one mentioned in the paper. Will check and update the repo for that if required by you guys. Sorry for the confusion.

[sourcesync]

Thanks @rahulk207. Just so you know...I'm currently running "main.py" and looks good so far. I'm just trying to reproduce your results on my machine before exploring some ideas around attention and transformer models. It would be nice to know what the gene expression numeric number means as well ( I'm not a domain expert in this area. ). Thanks for your follow-up!