In patterja/kbmtl: KBMTL Drug Prediction

Introduction

Workflow for Kernalized Semisupervised Classification Variational Method.

According to Gönen, Mehmet & Margolin, Adam A. 2014.Drug susceptibility prediction against a panel of drugs using kernelized Bayesian multitask learning. Bioinformatics >(Oxford, England) 30 (17): i556-563. doi:10.1093/bioinformatics/btu464.

Breast cancer cell line pre-processing

1) merge common cell lines between prediction and rnaseq 2) filter for protein coding genes only with targetid 3) log2 of training rnaseq 4) scale trainnig rnaseq 5) remove response with greater 0.95 and less than 0.05 frequences in both sens/resistant 6) expression genes with more than 0 expression in 1 sample, remove NAs 7) filter for trusight genes

save(state, file =  "smmart_trained_machine_learning_model.RData")
Y_predicted <- as.data.frame(prediction$P)
  colnames(Y_predicted) <- colnames(Y_train)
  return(Y_predicted, prediction)

patterja/kbmtl documentation built on Oct. 6, 2020, 1:16 p.m.