To DO
Age distribution comparison between POS and CONTROL
Current Plan
| Number | Task Description |
|---|---|
| 1. | Chromosome data download and processing |
| 2. | Generate the full Qnet-training-ready dataframe (SISA, PTSD, Uterine, Melanoma) |
| When we say Qnet-data, we mean the following: | |
| Tabular data, rows are samples, columns are ICD code (3 letters), table entries are the rest of the code, with 'H' if value is missing. This dataframe is generated once and is same for all problems. From this dataset, we generate a POS qnet specific to different problems. For a SI/SA POS qnet, we extract the subset of samples for which we have SI/SA up to 3 months before the target event, and construct Qnet. It is important to specify how we incorporate time. Maybe it can be the following format: + XXX_age_in_months. Or we can use: + XXX_time_to_end_of_record_in_months, where end of record is where we want the record to end... say 3 months before target event or 1 year before. If we do it this way, then the dataframe needs to be different for each problem instance. | |
| 3. | Incorporate PFSA processing in the ZBL pipeline |
| 4. | Incorporate SHAP analysis in the zcor pipeline to generate the individual patient SHAP CSVs |
| 5. | Need the complete and updated results for the current ZBL targets, with saved model objects. |
ZBL Targets
- ASD |
- ADRD |
- IPF |
- MACE |
- SI/SA |
- PTSD |
- CANCERS: Melanoma (with cpt), pancreatic, uterine, breast, lung, prostate, thyroid |
- BIPOLAR/MANIC