EMLI-ICC: An ensemble machine learning-based proteome and transcriptome integration algorithm for metastasis prediction and risk-stratification in intrahepatic cholangiocarcinoma

Intrahepatic cholangiocarcinoma (ICC) is the second-most-common primary liver cancer with increasing incidence and mortality worldwide. Although tumor metastasis is frequent even in ICC, improved robust strategies to identify patients at high risk for metastasis remains limited. Batch effects and different data types greatly decrease the predictive performances of signatures based on gene/protein expression profiles in inter-laboratory and different data type validation. To address this problem and assist in more precise diagnosis, we performed a genome-wide integrative analysis of proteome and transcriptome and developed an ensemble machine learning-based integration algorithm for metastasis prediction and risk-stratification in ICC (EMLI-ICC). Based on massive proteome and transcriptome data sets, 186 feature (biomarker) genes were selected and used to train the EMLI-ICC algorithm.The metastasis prediction model based on the EMLI-ICC algorithm showed AUC 0.887 for determining the metastasis samples in proteome and transcriptome datasets. To test prediction accuracy of our EMLI-ICC algorithm, we evaluated two RNA-seq datasets (TCGA, and Ahn’s data) and found AUC 0.923 and 0.883, respectively. We next analyzed 103 specimens from patients with ICC (metastasis, n = 55; non- metastasis, n = 48), followed by Cox proportional hazard regression analysis, to develop an integrated prognostic gene panel and establish a risk-stratification model for metastasis in ICC. We report an ensemble machine learning-based integration algorithm for metastasis prediction and risk-stratification that is superior to currently used clinicopathological features in patients with ICC.