Precision Medicine for All: Using Tidymodels to Validate PRS in Brazil
summary
Uses tidymodels to validate European-derived breast cancer PRS in a highly admixed Brazilian cohort (≈75% multi-ancestry), showing three PRSs tested, two validated (3h0 and 313; 3h0 performs best) and one failing in the East Asian–majority subgroup due to wide CIs. Workflow uses rsample, recipes, parsnip, workflows, yardstick for stratified splits, control-normalized PRS, logistic regression, and ROC/AUC; top PRS decile yields odds ratios comparable to ATM/CHEK2. Result becomes a deployed Mendelix genetic test including PRS and top 100 cancer genes.
AI-written summary · about
chapters (8)
- 0:00Precision medicine & PRS validation
- 3:56European PRSs in Brazil challenges
- 7:16Overcoming overwhelm with tidymodels
- 9:05Tidymodels workflow for PRS
- 11:17Validating PRSs across ancestries
- 13:34Clinical significance of PRS results
- 14:41Key takeaways on precision medicine
- 15:49Q&A
related talks
35mUsing Large-Scale Clinico-Genomics Data for in silico Clinical Trials and Precision Oncology
29mDeveloping and Delivering Personalized Polygenic Scores at Scale
22mInferring Selective Advantage Relationships to Reconstruct Cancer Progression Models
42mManualRivas
24mAugmenting machine learning with experimental data for drug...
50mThe algorithm for precision medicine
53mThe Algorithm for Precision Medicine
19mtidymodels for time-to-event data
