Screening Of PM_2.5 Controlling Pollutants And Its Lung Damage Effects Based On Machine Learning

With the development of industrialization,particulate matter pollution,mainly fine particulate matter(PM_2.5),has become one of the major environmental pollution problems in China,especially in northern China,where heating in winter makes particulate matter pollution worse.The composition of particulate matter is complex,and the identification of effective pollutants and their toxic damage is a hot topic in the field of environmental health.Therefore,it is necessary to systematically study the toxic components and effects of particulate matter.At present,a large number of relevant studies have explored the composition of particulate matter and its possible effects on health damage.However,these traditional experiments usually require a large amount of manpower and material resources,and the economic cost is high,so the data volume is not enough to represent all the characteristics of particulate matter.It is necessary to combine our previous research work and dig into the existing research deeply based on the open-source shared database.Computer techniques such as machine learning help to combine accumulated data from various similar or identical experiments for overall analysis.At present,a variety of machine learning models have been applied to the prediction of PM_2.5concentration changes,but there is still a lack of widespread application in the screening of main pollutants and characteristic genes after exposure to PM_2.5.Therefore,this study will be based on machine learning,combined with traditional biological information and toxicology analysis methods,to screen the main control pollutants and related target genes after PM_2.5exposure,and verify and discuss the gene expression and related health risk evaluation.1.This part of the study collected relevant literature on the influence of PM_2.5-exposed cells on cell activity(Pubmed was used to collect human non-small cell lung cancer cells（A549）and bronchial epithelial cells（Beas-2B）exposed to PM_2.5),and applied three models of machine learning,namely:Light GBM model,XGBoost model and Random Forest（RF）model were used to screen the characteristics of PM_2.5main pollutants affecting cell activity.Firstly,the collected literature and data were sorted out.According to cell type,the data sets were divided into whole cell data sets,cancer cell and non-cancer cell data sets,bronchial cell data sets,and lung parenchymal cell data sets.Three machine learning models were applied to screen core pollutant components respectively.The results showed that in the whole-cell data set,the influence of PM_2.5loading manganese（Mn）,zinc（Zn）,and lead（Pb）on cell activity was higher.Comparing the data sets of cancer cells and non-cancer cells,Pb,cadmium（Cd）,Zn,and Mn were mainly affected by cancer cells,while phenanthrene（PHE）and iron（Fe）were more toxic to non-cancer cells.In the data set of lung parenchymal cells,Fe,copper（Cu）,Mn,and Zn were the main pollutants affecting the cell activity of PM_2.5.In bronchial cell data sets,Mn,Fe,Benzo（a）pyrene（Ba P）,and naphthalene（NAP）were the main PM_2.5pollutants affecting cell activity.Therefore,it can be speculated that Mn,Fe,Zn,and other components may be the main load pollutants of PM_2.5affecting lung tissue cells.2.In this part of the study,the key expression genes of PM_2.5exposure-induced lung injury were further explored.First,a PM_2.5mouse exposure model was established.After exposure,mouse lung tissues were collected and gene microchip was performed.Then,the representative data sets related to lung injury in mice after PM_2.5exposure were further screened from the GEO public database.On this basis,machine learning and traditional bioinformatics analysis methods were used to calculate the differentially expressed genes in public data sets and experimental gene chip,and the differences and similarities of differentially expressed genes in different analysis methods were compared.Based on the above two methods,18 common genes were screened.Five of the genes can be converted into human genes.Survival analysis of the five genes showed that the group with high expression of all five genes had a greater risk of death.Through RNA extraction and real-time fluorescence quantitative PCR verification,it was found that the key target genes causing lung injury were PRL,MSC,Ssx9,Olfr1408,Olfr1322,Cd1d2,and PTPRN.In this study,a large number of existing literature were screened and summarized,and machine learning was used to screen Zn,Mn,and Fe,the main loading components of PM_2.5causing cytotoxicity.On this basis,public and experimental data sets were systematically analyzed to clarify the main target genes that may be involved in PM_2.5-induced lung injury.This study provides an experimental and theoretical basis for PM_2.5pollution control and health risk assessment.