First generate the tract profiles by following the steps outlined here: https://github.com/36000/OR_aging_ukbb

To process individual subjects, follow all of the instructions in the afq_processing folder of that repository.

To generate the tract profiles, follow the steps in the analysis_and_figures folder up to running gen_pheno.py .

You can also use the python_analysis_docker from that repository to run these scripts.

For some of these plots, you can also skip processing of all of the subjects, and instead use the aggregrate data provided in aggregate_data. For others, the models that generate the results are provided in ML_models.

For figure 1, run: first_paper_plots/general_profiles_plot.py glauc_match_profs_left

NOTE: The figure 1 script can also be run on profiles_w_aci_glauc_sec_match.csv in aggregate_data.

For figures 2 and 3, run primary_match.py with pheno_of_interest set to "glauc_sec"

NOTE: The models used to generate figures 2 and 3 are provided in ML_models under glaucoma.

For figure 4, run these series of commands:

• primary_match.py with pheno_of_interest set to "age"
• secondary_match.py amd_rob
• secondary_match.py tenano
• secondary_match.py glauc_rob
• secondary_match.py amd_rob_age

NOTE: The models used to generate figure 4 are provided in ML_models under age.

For supplementary figures 1 and 2, run pheno_histograms.py

NOTE: Supplementary figures 1 and 2 require access to UK Biobank phenotypic data which cannot be made publicly available.

For supplementary figure 3, run first_paper_plots/general_profiles_plot.py glauc_match_profs_right

NOTE: This script can also be run on profiles_w_aci_glauc_sec_match.csv in aggregate data.

For supplementary figure 4, run missing_bias.py

NOTE: Supplementary figure 4 requires access to individual tract profiles, which cannot be made publicly available.

For supplementary figure 5, run full_bundle_test.py

NOTE: Supplementary figure 5 uses models provided in ML_models under glaucoma.

Notice that there are many options for pheno_of_interest in primary_match.py. This includes the options "glauc_sec_no_cut" and "age_nocut", which test whether the main results of the paper hold when not cutting subjects with high logmar values. It also includes options for doing statistical matching with other control variables, other cutoff values, and other techniques. In all cases, the main results of the paper hold.