Multi-omics analyses develop and validate the optimal prognostic model on overall survival prediction for resectable hepatocellular carcinoma.

Background: Prediction of prognosis in patients with hepatocellular carcinoma (HCC) by single-omics profiling has been widely studied. However, the prognosis related to biomarkers of multiple omics has not been investigated. We aimed to establish and validate a prediction model for prognosis prediction of resectable HCC combining multi-omics and clinicopathological factors.

Methods: The training cohort involved multi-omics data of 330 patients with resectable HCC (stage I-IIIA) at mutational, copy number variation (CNV), transcriptional, and methylation levels from The Cancer Genome Atlas (TCGA) database, along with clinicopathological information. The validation cohort involved samples from 40 HCC patients of Beijing Youan Hospital. Univariate and multivariate analyses were performed in single-omics with clinicopathological variables regarding patient prognosis, and independent risk factors were combined to establish the multi-omics model. The predictive accuracy was assessed by the receiver operating characteristic (ROC) method.

Results: The mutational, copy number, transcriptional, and methylation alterations in HCC were characterized. TP53, CTNNB1, and TTN were among the genes with the top mutational frequency, and FBN1 and MAP1B mutations were independent risk factors for patient overall survival (OS). 1q21.3 and 1q23.3 ranked the highest in copy number amplifications, and 8p12 and 8p23.3 ranked the highest in deletions, and CSMD1, TP53, and RB1 were genes with the most frequent CNVs. AFP, GPC3, and TERT were among genes with the most significant aberrant transcription, and the transcription of CCNJL, FRMD1, and GRPEL2 were independent risk factors for OS. Both hypermethylation and hypomethylation can be observed. The aberrant methylation of CXorf15, DACT2, GP6, KIAA1522, and PDIA3 were independent risk factors. Single-omics models were established with independent risk factors, and were validated by internal and external datasets. A prognostic model for OS with multi-omics independent risk factors and clinicopathlogical information was established. Internal and external validation achieved an optimal maximal area under the curve (AUC) of 0.98 at 1 year and 0.88 at 2 years, respectively.

Conclusions: A multi-omics model combining molecular aberrancies and clinicopathological information was established and proved to be optimal for prognosis prediction of resectable HCC. This model may be helpful for therapeutic strategy selection and survival assessment.