Skip to main content

Table 2 Algorithm of the stepwise statistical analysis perfomed on the discovery and validation datasets

From: Whole-blood RNA transcript-based models can predict clinical response in two large independent clinical studies of patients with advanced melanoma treated with the checkpoint inhibitor, tremelimumab

Step-Wise Statistical Analysis Using Discovery/Validation Methodology
Step 1: 2-Gene Models to Predict Immunotherapy Response and Survival
 CORExpress 1.1 regression analysis software for high-dimensional data
 Train 2-gene models with pre-treatment Discovery dataset N = 210
 Test 2-gene models with pre-treatment Validation dataset N = 150
 Over 260 2-gene synergistic pre-treatment models trained and validated
Step 2: Larger Gene Models to More Accurately Predict Response and Survival
 Include only genes validated in 2-gene models from Step 1
 Optimize model coefficients using CORExpress 1.1 software
 Train optimized models with pre-treatment Discovery dataset N = 210
 Test optimized models with pre-treatment Validation dataset N = 150
Step 3: Finalize 15-Gene Model to Predict Response and Survival
 Optimal 15-gene pre-treatment model selected from Step 3
 Use MedCalc version 17 software for ROC and p-value analysis
 Test 15-gene model with pre-treatment Discovery dataset N = 210
 Test 15-gene model with pre-treatment validation dataset N = 150
Step 4: Test Pre-treatment 15-Gene Model with Post-Treatment Datasets
 15-gene pre-treatment response and survival model from Step 3
 Use MedCalc version 17 software for ROC and p-value analysis
 Test 15-gene model with post-treatment Discovery dataset N = 210
 Test 15-gene model with post-treatment Validation dataset N = 150