Survival Time Prediction from Mutation Profiles Using Gene Networks

on YouTube ↗15m

" summary

Uses TCGA exome binary matrices and models survival with a linear model learned by ranking SVM, reporting concordance around 0.5–0.55 for raw mutation profiles. Applying network propagation plus quantile normalization (NSQn) raises performance; a simpler two-step algorithm called NetNorm (ranking genes by degree, then normalizing top k to 1) matches NSQn. Real gene networks give significantly better predictions than randomized ones (t-test). One-iteration propagation still helps, and neighbor mutational burden plus normalization prove crucial for stratification and improved survival signals.

AI-written summary · about

" chapters (6)

1. 0:00Start
2. 0:21Assessing the Prognostic Power of Mutation Profiles
3. 0:58Example of Survival Predictions with Tcga Exome Sequencing Data
4. 1:47The Concordance Index
5. 2:29Patient Stratification
6. 7:31Ranking and Normalization