Comprehensive machine-learning survival framework develop a consensus model in large scale multi-center cohorts for pancreatic cancer

This article has 3 evaluations Published on
Read the full article Related papers
This article on Sciety

Abstract

Background

As the most aggressive tumor, the outcome of pancreatic cancer (PACA) has not improved observably over the last decade. Anatomy-based TNM staging does not exactly identify treatment-sensitive patients, and an ideal biomarker is urgently needed for precision medicine.

Methods

A total of 1280 patients from 10 multi-center cohorts were enrolled. 10 machine-learning algorithms were transformed into 76 combinations, which were performed to construct an artificial intelligence-derived prognostic signature (AIDPS). The predictive performance, multi-omic alterations, immune landscape, and clinical significance of AIDPS were further explored.

Results

Based on 10 independent cohorts, we screened 32 consensus prognostic genes via univariate Cox regression. According to the criterion with the largest average C-index in the nine validation sets, we selected the optimal algorithm to construct the AIDPS. After incorporating several vital clinicopathological features and 86 published signatures, AIDPS exhibited robust and dramatically superior predictive capability. Moreover, in other prevalent digestive system tumors, the 9-gene AIDPS could still accurately stratify the prognosis. Of note, our AIDPS had important clinical implications for PACA, and patients with low AIDPS owned a dismal prognosis, relatively high frequency of mutations and copy number alterations, and denser immune cell infiltrates as well as were more sensitive to immunotherapy. Correspondingly, the high AIDPS group possessed dramatically prolonged survival, and panobinostat might be a potential agent for patients with high AIDPS.

Conclusions

The AIDPS could accurately predict the prognosis and immunotherapy efficacy of PACA, which might become an attractive tool to further guide the stratified management and individualized treatment.

Funding

This study was supported by the National Natural Science Foundation of China (No. 81870457, 82172944).

Related articles

Related articles are currently not available for this article.