Skip to main content
Fig. 3 | BMC Pulmonary Medicine

Fig. 3

From: A functional genomic model for predicting prognosis in idiopathic pulmonary fibrosis

Fig. 3

Genomic model and 10-fold cross validation results. a A genomic model was constructed from the 118 IPF prognostic predictor genes using “Survival risk group prediction” algorithm implemented in BRB-ArrayTools (see Additional file 1) followed by 10-fold cross validation (CV) algorithm to calculate the misclassification rate. Formula of genomic model: Prognostic index (PI) = ∑W i * X i +13.5, where W i and X i represent the weight and log-intensity of the i-th gene in IPF prognostic predictor gene set compiled from training cohort, respectively. Misclassification rate (20 %) was determined by 10-fold CV and computed as (k + n)/total cases, where k represents the predicted high risk that are observed as good prognosis, and n represents the predicted low risk that are observed as poor prognosis. b IPF patients with predicted low (dotted line) and high risk (dashed line) stratified by prognostic index (PI) derived from each patient in training cohort based on the genomic model. The red line denotes 50 % probability of survival. PI independently predicted survival in univariate competing-risk Cox regression (Sub-hazard ratio (SHR) 2.7; 95 % CI 1.9-3.9; p < 0.001) and in multivariate competing-risk Cox regression after adjustment for baseline CPI (SHR 2.3; 95 % CI 1.5-3.4; p < 0.001)

Back to article page