Bidirectional deep neural networks to integrate RNA and DNA data for predicting outcome for patients with hepatocellular carcinoma.

ABSTRACT

Aims: Individualized patient profiling is instrumental for personalized management in hepatocellular carcinoma (HCC). This study built a model based on bidirectional deep neural networks (BiDNNs), an unsupervised machine-learning approach, to integrate multi-omics data and predict survival in HCC. Methods: DNA methylation and mRNA expression data for HCC samples from the The Cancer Genome Atlas database were integrated using BiDNNs. With optimal clusters as labels, a support vector machine model was developed to predict survival. Results: Using the BiDNN-based model, samples were clustered into two survival subgroups. The survival subgroup classification was an independent prognostic factor. BiDNNs were superior to multimodal autoencoders. Conclusion: This study constructed and validated a BiDNN-based model for predicting prognosis in HCC, with implications for individualized therapies in HCC.

Lay abstract In this study, one unsupervised machine-learning algorithm, namely bidirectional deep neural networks (BiDNNs), was used to learn DNA methylation data and RNA-seq data for hepatocellular carcinoma (HCC) patients, and correlations were captured. Based on the output of BiDNNs, HCC patients were classified into a good prognosis subgroup and a poor prognosis subgroup using the K-means clustering method. Patients in the first group are more likely to survive or live much longer than patients in the second group. The survival subgroups were closely associated with the survival of HCC patients. Different treatment options should be chosen for the two subgroups. This study proposes a BiDNNs-based strategy for survival stratification in HCC and may help doctors select optimal therapies for each patient.