Function-driven multi-molecular marker for individualized prediction of colorectal cancer prognosis and device and evaluation method thereof

Abstract

The invention discloses a function-driven multi-molecular marker for individualized prediction of colorectal cancer prognosis and a device and an evaluation method thereof. Prognostic predictive markers are simultaneously screened from multiple cancer large sample data sets of multiple centers, and predictive effects of the multi-molecular marker are validated and comprehensively evaluated in three independent data sets. The application of the above methods and strategies ensures the reliability of the multi-molecular marker in judgment of the prognosis of a colorectal cancer patient; the multi-molecular marker combination is not affected by the batch effect of an experiment or the difference of a detection platform, thereby ensuring the stability of the multi-molecular marker in the judgment of the prognosis of the colorectal cancer patient; the multi-molecular marker combination does not need to be subjected to data standardization between multiple samples before use, thereby facilitating the use and ensuring the clinical operability of the multi-molecular marker in the judgment of the prognosis of the colorectal cancer patient.

Description

technical field [0001] The invention relates to the technical fields of medical genomics and oncology, in particular to a function-driven individualized multi-molecular marker for predicting the prognosis of colorectal cancer, a device and an evaluation method thereof. Background technique [0002] Colorectal cancer is one of the most common malignant tumors in the world. With economic development and changes in eating habits, the incidence of colorectal cancer is increasing year by year, and its mortality rate ranks fifth among cancer-related deaths in China. Globally, colorectal cancer ranks fourth among male malignant tumor deaths and third among females. Colorectal cancer mortality is higher in developing countries (approximately 52% of colorectal cancer deaths), indicating that colorectal cancer survival rates in these regions are very low. Globally, 1.2 million new cases are diagnosed each year, and more than 600,000 patients die from colorectal cancer. The incidence...

AI Technical Summary

Problems solved by technology

[0003] Although some studies have proposed some molecular markers for the prognosis of colorectal cancer, these studies are mainly based on a single data set, a single technology platform or a small sample data set, so the established cancer prognosis model is usually too dependent on the platform where the sample is located. Technical characteristics, with a certain degree of bias, often lead to overfitting, and it is difficult to obtain robust molecular markers, so it is impossible to accurately predict the prognosis of colorectal cancer patients in different platform data

At the same time, the existing molecular marker combinations often include a large number of genes, and the repetition rate between different molecular marker combinations is small, which is not suitable for clinical application detection

Finally, the existing molecular marker combinations are often based on the linear scoring model of the training set, and the scoring threshold is set. However, when the scoring model is applied to an independent data set, the data set needs to be standardized, and it is impossible to perform individualized risk prediction for a single patient.

Therefore, the combination of molecular markers discovered in the past is relatively one-sided in predicting and evaluating the prognosis of colorectal cancer, and the sensitivity and specificity are often not ideal.

Images