基于人工智能技术的烟气暴露大鼠代谢生物标志物筛选方法研究

Study on Screening of Cigarette Smoke Exposure Biomarkers for Rat's Metabolites on the Basis of Artificial Intelligence Technologies

该研究将主成分分析、偏最小二乘判别分析等多元统计分析方法用于烟草血浆、尿液和肺组织代谢组学数据的分析,以揭示暴露于不同烟气中大鼠血浆、尿液和肺组织中内源性生物标志物的整体变化情况,筛选潜在生物标志物;将血样、尿样和肺组织代谢轮廓谱分析得到的生物标志物进行整合,运用神经模糊网络模型对标志物进行缩减,并用人工神经网络评价模型预测能力,确定烟气暴露不同时间(7,14,30 d)以及不同烟气暴露对大鼠内源性代谢物变化影响"因果效应"密切相关的关键生物标志物群,明确不同烟气对大鼠机体损伤机制的异同。

Multivariate statistical analysis methods, principal component analysis and partial least square discrimination analysis, were applied in this study for the data mining of cigarette smoke expo- sure metabolomics on plasma, urine and lung samples, in order to characterize the holistic influences of cigarette smoke exposure, and screen potential biomarkers. The screened biomarkers obtained from the metabolic profiling analysis on plasma, urine and lung were integrated and reduced by neu- rofuzzy logic. The predictability of the established model with this focused biomarkers were evaluated by artificial neural networks. Key biomarkers were closely related to different smoke exposure time （7, 14, 30 days） , and different kinds of cigarette smoke exposure on the endogenous metabolites in rats were found in this study, and the damage mechanism of cigarette smoke exposure on rat＇s organ- ism was discussed.