Skip to content

CDA Disease

We extract information about the disease diagnosis from two CDA tables, diagnosis and researchsubject. We first summarize the tables and then outline our ETL strategy.

diagnosis

Column Example Explanation
diagnosis_id CGCI-HTMCP-CC.HTMCP-03-06-02424.HTMCP-03-06-02424_diagnosis y
diagnosis_identifier see below y
primary_diagnosis Squamous cell carcinoma, keratinizing, NOS y
age_at_diagnosis 13085.0 y
morphology 8071/3 y
stage None y
grade G3 y
method_of_diagnosis Biopsy y
subject_id CGCI.HTMCP-03-06-02424 y
researchsubject_id CGCI-HTMCP-CC.HTMCP-03-06-02424 y

The fields of the table have the following meaning.

  • diagnosis_id Question: It seems as if this identifier has some syntex of meaning or is it random?
  • diagnosis_identifier Question: This field seems to have a lot of structure. How is it used in CDA and is there documentation on how to interpret it? This field has the following structure. 
    [{'system': 'GDC',
  'field_name': 'case.diagnoses.diagnosis_id',
  'value': '06af070e-aad4-5b2d-a693-b6ccfe93985a'},
 {'system': 'GDC',
  'field_name': 'case.diagnoses.submitter_id',
  'value': 'HTMCP-03-06-02424_diagnosis'}]
  • primary_diagnosis This field represents the main cancer diagnosis of this individual
  • age_at_diagnosis This field represents the number of days of life of the individual on the day during which the cancer diagnosis was made.
  • morphology Entries such as 8071/3 are ICD-O codes. TODO - translate into ontology codes.
  • stage Cancer stage.
  • grade Cancer grade. Note that in many tables there are strings such as G3. NCIT has more detailed terms, but we think it best to stick to the top level, and possible consider postcomposition to represent specific stage systems.
  • method_of_diagnosis This corresponds to
  • subject_id Identifier for the individual being investigated
  • researchsubject_id Identifier for the researchsubject (which can be a sample or an individaul - Question: where is this documented?)

researchsubject

Column Example Explanation
researchsubject_id CPTAC-3.C3L-00563 y
researchsubject_identifier see below y
member_of_research_project CPTAC-3 y
primary_diagnosis_condition Adenomas and Adenocarcinomas y
primary_diagnosis_site Uterus, NOS y
subject_id CPTAC.C3L-00563 y
  • researchsubject_id xyz

  • researchsubject_identifier Question: How do we interpret this kind of structure: 

    [{'system': 'GDC',
  'field_name': 'case.case_id',
  'value': '2b1894fb-b168-42ca-942f-a5def0bb8309'},
 {'system': 'GDC', 'field_name': 'case.submitter_id', 'value': 'C3L-00563'}]

  • member_of_research_project Question: Where do we get more information about the research projects? What informationis available?

  • primary_diagnosis_condition Question: This seems to be duplicative with the field primary_diagnosis in the diagnosis table. What is the difference?
  • primary_diagnosis_site Todo - we can map this to uberon
  • subject_id This relates to the subject_id in other tables.

Mapping strategy

We merge the diagnosis and researchsubject tables to retrieve all needed information about the disease diagnosis.

Merging diagnosis and researchsubject via the researchsubject_id
merged_df = pd.merge(diagnosis_df,
                    rsub_df,
                    left_on='researchsubject_id',
                    right_on='researchsubject_id',
                    suffixes=["_di", "_rs"])