---

---

---

TB_RNA_Seq: Merging Datasets + Assignment Trees

There are three R scripts in this project

generate_key.R: It merges two Redcap Databases; the smaller RNA seq study (NCR Database) with the larger GWAS study (NCTB Database) using fuzzy name matching, saliva barcodes and manual matching. I

case_decision_tree.R: Assigns individuals as cases or "other" based on clinical symptoms, initial self-report and verified with current medical data

control_decision_tree.R :Assigns individuals as controls based on PRIOR medical records. Searches a wide range of variables for TB evidence in their lifetime.

Note that cases and control are assigned in separate scripts due to different criteria from different databases used for assignments

Before running Scripts:

Before running, you must

1) In each script manually check that the data file is the most recent from REDCAP, if not re-download. The variables in the scripts where you edit these are:

a) control_decision_tree.R : raw_NCR_data

b) generate_key.R: raw_NCTB_data & raw_NCR_data

c) case_decision_tree.R: raw_NCR_data_

2) In updated_data_merge.R you must change output_path to your actual desired path on dropbox or your desktop (it's Oshi's by default)

Running Scripts:

There are 2 options to run the scripts

The scripts are run in the folllowing order

Generate key -> control_decision_tree | case_decision_tree

1. If you just want the NCTB-NCR key: just run generate_key.r

2. To get a list of validated case or controls run their respective decision tree. This automatically sources generate_key.r. . You should get an output pop up on your rstudio with final case/control counts and also a larger merged table with the variables used to assign cases and controls

3) Investigator mode: You would do this when trying to find cracks in the system like when patients violate any of the assignment criteria and why. Remember You can run case_decision_tree or control_decision_tree in any order.

After running case_decision_tree :

• view(Validatedcases2) and check out these columns for individuals that have become "notcase_other" :

1. lab_date_2_diag_date = lab_receive_date - TB_Diag_Date,

2. lab_date_2_trtment_strt = lab_receive_date - treatment_srt_date,

3. less_than_thirty = lab_receive_date - TB_Diag_Date < 30,

4. intermediate2 = tb_status == "Yes",

5. intermediate3 = HIV_status == "Negative"

   if any of these columns are False or >15 or 30 then that provides the context

control_decision_tree:

view(final_control_validated_tbl)

check notctrl_other and their corresponding val_sr and val_med_tb columns

if val_sr is false:

• view(self_report_validated2) and check prior_tb_self_reported, prior_n_tb_self_reported, frst_pst_tb_test_date

if val_med_tb is false;

• check tb_test and treat_or_drug

#to do: make another script from updated data_merge to check for missing demographic data