620P - A unique circulating microRNA pairs signature serves as a superior tool for early diagnosis of pan-cancer

Background

Cancer constitutes a major burden to global health and the critical role of early diagnosis for cancer management is self-evident. Even though various miRNA-based signatures have been developed, their clinical utilization is limited due to various reasons. In this article, we innovatively developed a signature based on pairwise expression of miRNAs (miRPs) for pan-cancer diagnosis using machine learning approach.

Methods

miRNA spectrum of 15832 patients with 13 different cancers from 10 cohorts were analyzed. 15148 patients were divided into training, validation, and test sets with a ratio of 7:2:1, while 648 patients were utilized as external test. Pairwise comparison was performed to generate miRP score, defined by the comparison between two miRNAs, in training set. Five different machine-learning (ML) algorithms (XGBoost, SVM, RandomForest, LASSO, and Logistic) were adopted for signature construction. The best ML algorithm and the optimal number of miRPs included were identified using AUC and youden index in validation. Performance of the ideal model was evaluated in test and external set based on AUC, Youden index, positive predictive value (PPV), negative predictive value (NPV), sensitivity, specificity, and accuracy. The AUC of entire cohorts was compared to previously published 25 signatures.

Results

The Random Forest approach including 31 miRPs (31-miRP) outperformed others and was retained for further evaluation. The AUC of 31-miRP ranges 0.980-1.000 in different set. Remarkably, 31-miRP exhibited advantages in differentiating different cancers from normal tissues. Moreover, 31-miRP demonstrate superiorities in detecting early-stage cancers, with AUC ranging from 0.961-0.998. Compared to previously published 25 different signatures, 31-miRP also demonstrated clear advantages. Remarkably, 31-miRP also exhibited promising capabilities in differentiating cancers from corresponding benign lesions.

Conclusions

The 31-miRP exhibited outstanding diagnostic performance, characterized by high accuracy and sensitivity, thereby holding potential as a reliable tool for cancer diagnosis at early stage. Nevertheless, its effectiveness still warrants further investigation in real-world setting in future.

Clinical trial identification

Editorial acknowledgement

Legal entity responsible for the study

The authors.

Funding

CAMS Innovation Fund for Medical Sciences (No.2021-I2M-1-050); National Natural Science Foundation for Young Scientists of China (No. 82203025).

Disclosure

All authors have declared no conflicts of interest.