Supplementary Material for “A Bayesian hierarchical model uncovers targetable oncogenic hotspots in disordered regions of FGFRs” by Hommerich, Naveja, et al.

Table S1. Samples from UCT Mainz cancer patients with FGFR mutations.
Table S2. Comparison of the area under the ROC curve for the different models
predicting the oncogenic effect of 137 missense mutations as reported by Nakamura et al.
Table S3. We extracted the clinical results from RAGNAR for 22 missense mutations included
there and scored them using our model. For the response rate we compute the proportion of
responders (either CR or PR). Our prediction is based on whether we find the residue position
to be a hotspot with at least 99% probability.
Table S4. Mutations reported in the UCT dataset and whether they are included in RAGNAR or
predicted to be oncogenic by BRIGHT-FGFR.
Table S5. Alpha estimates from the model and statistics of the posterior distribution. MAP:
mean a posteriori. aa_pos: amino acid position. P: percentile of the posterior distribution. 
Table S6. Beta (BRIGHT-FGFR) estimates from the model and statistics of the posterior
distribution. MAP: mean a posteriori. aa_pos: amino acid position. P: percentile of the posterior
distribution.