Fig. 6. Prediction model construction using TTN-TMB and association between immunostimulatory signature and mutation.

a Overall process of constructing the prediction model to classify MSI-H and MSS by machine learning. b Internal and external validation of the diagnostic model for MSI-H constructed using Random Forest machine learning in STAD, UCEC, and CRC. c Enrichment scores associated with immunostimulatory signature (IS) for each gene mutation. TTN mutations were significantly enriched in samples possessing high IS score (permutated p < 0.0001; 10,000 random sample class permutations). d IS score was significantly higher in the TTN mutated group than in the TTN wild-type group (p < 2.2e-16, Wilcoxon rank-sum test). AUC area under the curve, N total mutation number.