--- a
+++ b/data_prep/README.md
@@ -0,0 +1,41 @@
+# Prepare your own patient dataset
+
+## Preprocessing steps
+Your preprocessing script must do the following:
+1. Map genes to Ensembl IDs
+2. Map phenotypes to the 2019 version of HPO
+3. Output a jsonlines file where each json (i.e., line in the file) contains information for a single patient
+
+Please refer to the `create_mygene2_cohort/preprocess_mygene2.py` for an example preprocessing script.
+
+## Patient information
+
+An example patient from the simulated patients dataset:
+
+```
+{
+ "id": 9,
+ "positive_phenotypes": ["HP:0000221", "HP:0000232", "HP:0001155", "HP:0005692", "HP:0012471", "HP:0100540", "HP:0001999", "HP:0001249", "HP:0010285", "HP:0000924", "HP:0004459"],
+ "all_candidate_genes": ["ENSG00000196277", "ENSG00000104899", "ENSG00000143156", "ENSG00000088451", "ENSG00000157557", "ENSG00000165125", "ENSG00000157766", "ENSG00000108821", "ENSG00000142655", "ENSG00000184470", "ENSG00000157119", "ENSG00000069431", "ENSG00000131828", "ENSG00000179111", "ENSG00000168646"],
+ "true_genes": ["ENSG00000069431"],
+ "true_diseases": ["966"]
+}
+```
+
+### Required
+
+The minimal information required for each patient are:
+- Patient ID ("id")
+- List of phenotypes present in the patient as HPO terms ("positive_phenotypes")
+
+To run causal gene discovery, the json must also include:
+- List of all candidate genes as Ensembl IDs ("all_candidate_genes")
+
+To run patients-like-me identification or novel disease characterization, the json does not require any additional information.
+
+### Optional
+- Causal genes ("true_genes"). *If available, please provide causal genes as Ensembl IDs.*
+- Disease names ("true_diseases"). *If available, please provide true disease names as MONDO IDs.*
+- Omim ID
+- Orphanet ID
+- Orphanet category