基于随机森林与人工神经网络慢性鼻窦炎伴鼻息肉诊断模型的构建与分析

Construction and analysis on the diagnostic model of chronic rhinosinusitis with nasal polyps based on random forest and artificial neural network

摘要:

目的 慢性鼻窦炎伴鼻息肉(CRSwNP)是一个全球性的健康问题，现有的诊断技术存在一定局限性，因此有必要开发新的诊断模型来补充现有的诊断方法。方法 利用CRSwNP患者(GSE23552、GSE36830)的公开基因表达数据来识别潜在的差异基因，应用随机森林算法和人工神经网络进一步筛选特异性基因，建立CRSwNP的早期诊断模型。结果 共发现78个上调基因和25个下调基因，随机森林算法筛选了4个特异性基因(HPGDS、IL1RL1、FMO3、DOK3)，人工神经网络构建出基于上述基因的预测模型，该模型具有良好的预测效果(AUC=0.986)，独立数据集GSE194282进一步验证了其准确性(AUC=0.888)。结论 采用机器学习方法建立了一个基于基因表达水平的预测模型，该模型可以预测早期CRSwNP，有助于早期诊断和改善临床决策。

关键词:慢性鼻窦炎伴鼻息肉;诊断模型;随机森林;人工神经网络

Abstract:

Objective Chronic rhinosinusitis with nasal polyps (CRSwNP) is a global health problem, and the existing diagnostic techniques have some limitations. Therefore, it is necessary to develop new diagnostic models to supplement the existing diagnostic methods.Methods The public gene expression data of CRSwNP patients (GSE23552, GSE36830) were used to identify potential differential genes. The random forest algorithm and artificial neural network were used to screen specific genes and establish the early diagnosis model of CRSwNP.Results A total of 78 up-regulated genes and 25 down-regulated genes were identified. Four specific genes (HPGDS, IL1RL1, FMO3 and DOK3) were screened by random forest algorithm. The prediction model based on the above genes was constructed by artificial neural network, which had good prediction effect (area under the curve=0.986). Independent dataset GSE194282 further verified the accuracy (area under the curve=0.888).Conclusions A predictive model based on gene expression level is established by machine learning method. This model can predict early CRSwNP, which is helpful for early diagnosis and clinical decision.

Key words:Chronic rhinosinusitis with nasal polyps;Diagnostic model;Random forest;Artificial neural network