stuppie
commented
7 years ago @andrawaag @andrewsu @sebotic @lschriml I sketched out a single example to use to discuss a few things. This is only one example and there will be other issues with other entries.... There's no way to link to a single entry, so ho here: https://www.cancergenomeinterpreter.org/biomarkers and type "BRCA1 deletion" in the "Biomaker" box. The data from the tsv file download is below:
| Alteration | Alteration type | Assay type | Association | Biomarker | Curator | Drug | Drug family | Drug full name | Drug status | Evidence level | Gene | Metastatic Tumor Type | Primary Tumor acronym | Source | Targeting | individual_mutation | transcript | gene | strand | region | info | cDNA | gDNA | Primary Tumor type |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| BRCA1:del | CNA | Responsive | BRCA1 deletion | RDientsmann | [] | [PARP inhibitor] | PARP inhibitors | Pre-clinical | BRCA1 | OV | PMID:22392482 | Ovary |
1) Sometimes a drug family is specified, not a drug. So we need to be able to normalize all of these drug families. There are 116 unique drug family strings. I think most can be matched up to ChEBI by hand (example: PARP Inhibitors). We also need to be able to handle dual/multiple inhibitors (e.g. PI3K & MEK inhibitors ).
2) I'm not sure how to handle the evidence. CGI gives an "Evidence level", which may be one of the following: Pre-clinical, Early trials, Case report, FDA guidelines, European LeukemiaNet guidelines, NCCN guidelines, Late trials, CPIC guidelines, Clinical trial. In civic, this is a little more detailed because the evidence is specifically about the claim made by the evidence, but this is information about the source itself, and so I think it makes sense to put it on "stated in"'s item.
Thoughts?
sebotic
See https://www.cancergenomeinterpreter.org/biomarkers, a CC0 licensed database. Perhaps to be aligned with the CIVIC bot? cc @andrawaag Reference: http://www.biorxiv.org/content/early/2017/05/20/140475