背景:钉螺是日本血吸虫的唯一中间宿主。它的出现和复发对我国血吸虫病的消灭提出了持续的挑战。准确预测钉螺分布对血吸虫病防治具有重要意义。
方法:2016年湖北猪的分布数据来源于云南省地方病预防控制所。八种机器学习算法,包括极限梯度提升(XGB),支持向量机(SVM),随机森林(RF),广义增强模型(GBM),神经网络(NN),分类和回归树(CART),k-最近邻(KNN),和广义加法模型(GAM),被用来探索气候的影响,地理,和社会经济变量对湖北适宜地区分布的影响。对不同时期湖北适宜地区的分布进行了预测(2030年代,2050年,和2070)在不同的气候情景下(SSP126,SSP245,SSP370和SSP585)。
结果:RF模型表现出最佳性能(AUC:0.991,灵敏度:0.982,特异性:0.995,κ:0.942),CART模型表现最差(AUC:0.884,灵敏度:0.922,特异性:0.943,κ:0.829)。基于RF模型,前六个重要变量如下:Bio15(降水季节性)(33.6%),年平均降水量(25.2%),Bio2(平均昼夜温度范围)(21.7%),Bio19(最冷季度的降水)(14.5%),人口密度(13.5%),和夜光指数(11.1%)。结果表明,在当前的气候条件下,湖北的总体适宜生境主要分布在云南省西北部的血吸虫病流行区,并且由于气候变化,预计将向北和向西扩展。
结论:这项研究表明,对O.hupensis当前分布的预测与实际记录非常吻合。此外,我们的研究提供了令人信服的证据,表明蜗牛的地理分布预计将在未来几十年内向云南省的北部和西部扩展。这表明蜗牛的分布是由气候因素驱动的。我们的发现对于制定有效的蜗牛控制策略具有重要意义。
BACKGROUND: Oncomelania hupensis is the sole intermediate host of Schistosoma japonicum. Its emergence and recurrence pose a constant challenge to the elimination of schistosomiasis in China. It is important to accurately predict the snail distribution for schistosomiasis prevention and control.
METHODS: Data describing the distribution of O. hupensis in 2016 was obtained from the Yunnan Institute of Endemic Disease Control and Prevention. Eight machine learning algorithms, including eXtreme Gradient Boosting (XGB), support vector machine (SVM), random forest (RF), generalized boosting model (GBM), neural network (NN), classification and regression trees (CART), k-nearest neighbors (KNN), and generalized additive model (GAM), were employed to explore the impacts of climatic, geographical, and socioeconomic variables on the distribution of suitable areas for O. hupensis. Predictions of the distribution of suitable areas for O. hupensis were made for various periods (2030s, 2050s, and 2070s) under different climate scenarios (SSP126, SSP245, SSP370, and SSP585).
RESULTS: The RF model exhibited the best performance (AUC: 0.991, sensitivity: 0.982, specificity: 0.995, kappa: 0.942) and the CART model performed the worst (AUC: 0.884, sensitivity: 0.922, specificity: 0.943, kappa: 0.829). Based on the RF model, the top six important variables were as follows: Bio15 (precipitation seasonality) (33.6%), average annual precipitation (25.2%), Bio2 (mean diurnal temperature range) (21.7%), Bio19 (precipitation of the coldest quarter) (14.5%), population density (13.5%), and night light index (11.1%). The results demonstrated that the overall suitable habitats for O. hupensis were predominantly distributed in the schistosomiasis-endemic areas located in northwestern Yunnan Province under the current climate situation and were predicted to expand north- and westward due to climate change.
CONCLUSIONS: This study showed that the prediction of the current distribution of O. hupensis corresponded well with the actual records. Furthermore, our study provided compelling evidence that the geographical distribution of snails was projected to expand toward the north and west of Yunnan Province in the coming decades, indicating that the distribution of snails is driven by climate factors. Our findings will be of great significance for formulating effective strategies for snail control.