Supplemental data 1 for Gymrek et al: Interpreting short tandem repeat variations in humans using mutational constraint # Description of data file Column names are given in the comment line. Below is a brief description of each column: 1. chrom: chromosome name 2. start: start coordinate of the STR (according to lobSTR's hg19 reference) 3. end: end coordinate of the STR (according to lobSTR's hg19 reference) 4. est_logmu_ml: Estimated log10 mutation rate 5. est_beta_eff_ml: Estimated effective length constraint (beta/strsd**2) 6. est_beta_ml: Estimated length constraint 7. est_pgeom_ml: Estimated step size parameter 8. stderr_ml: Standard error on the mutation rate estimate. Negative standard errors are undefined. 9. numsamples_ml: Number of samples used in mutation rate estimation 10. up: Estimated probability for stutter to increase the unit length 11. down: Estimated probability for stutter to decrease the unit length 12. p: Estimated stutter step size parameter 13. period: length of the repeat unit 14. motif: motif of the repeat unit 15. uninterrupted_length: Longest interrupted perfect repeat track 16. pred_mu_1: Mu predicted by the neutral mutation model 17. pred_mu_se_1: standard error of pred_mu_1 18. zscore_1: Zscore calculated from pred_mu_1, pred_mu_se_1, est_logmu_ml, and stderr_ml 19. pred_mu_2: columns 18-20 are same as columns 15-17, but the model is trained without filtering loci with high or undefined standard errors 20. pred_mu_se_2 21. zscore_2 22. filter1: True if the locus has characteristics that were filtered before training Model 1 (columns 16-18, used in the main text). 23. train_1: True if the locus was included in training for Model 1. 24. train_2: True if the locus was included in training for Model 2. 25. smm_mu: log10 effective mutation rate estimated using a stepwise model 26. zscore_1_smm: Zscore calculated using mutation rates from the SMM model 27. filter1_smm: True if the locus was filtered before training the SMM-based constraint model # Notes on predicted mutation rates: Loci with NaN for fields 16-22 were not included in the constraint analysis. The main text reports constraint scores calculated using Model 1. The differences between Model 1 and Model 2 are as follows: Model 1: Loci with high, zero, or undefined standard errors were excluded from training (see column "filter1"). Model 2: Training did not exclude loci based on standard errors.