摘要
根据2021年5-8月淡水养殖池塘3个平行位点逐月实测数据,文章构建了基于支持向量机(SVM)的产毒微囊藻SNP/InDel基因型及胞内外毒素含量的预测模型。分析了各水化指标与产毒微囊藻6种基因型及胞内外毒素含量的相关系数,对原始数据归一化,以径向基函数为核函数,通过5折交叉验证确定模型最优参数,建立了用水化指标预测产毒微囊藻6种基因型的6个模型,以及用水化指标、6种基因型预测胞内外毒素含量的2个模型。结果表明:每项水化指标都至少与1种产毒微囊藻基因型或1项毒素含量的相关系数大于0.3。在8个预测模型中,对训练集预测的均方误差均小于0.07,对训练集预测的决定系数均大于0.9,平均的测试集均方误差为0.054。可见,利用SVM预测养殖池塘中产毒微囊藻SNP/InDel基因型及胞内外毒素含量是可行的,且精度较高。
Based on the monthly data measured at three parallel sites in freshwater aquaculture ponds from May to August 2021,a prediction model for SNP/InDel genotypes of toxic Microcystis and intracellular and extracellular toxin contents was established using support vector machine (SVM).The correlation coefficients between hydration indexes and 6 toxic Microcystis genotypes as well as intracellular and extracellular toxin contents were analyzed.The raw data were normalized,the radial basis function was used as the kernel function,and the optimal parameters of the models were determined by 5 fold cross validation.Six models were established to predict 6 toxic Microcystis genotypes using hydration indexes,and two models were established to predict intracellular and extracellular toxin contents using hydration index and 6 genotypes.The results showed that each hydration index had a correlation coefficient greater than 0.3 with at least one toxic Microcystis genotype or one toxin content.Among the 8 prediction models,the mean square deviations of the training set prediction were all less than 0.07,the coefficients of determination (R~2) of the training set prediction were all greater than 0.9,and the mean square error of test set was 0.054.It is feasible and accurate to predict SNP/InDel genotypes and intracellular and extracellular toxin contents in cultured ponds using SVM.
作者
刘琦
卫鹏
戴伟
毕相东
马志宏
LIU Qi;WEI Peng;DAI Wei;BI Xiangdong;MA Zhihong(Tianjin Key Laboratory of Aqua-ccology and Aquaculture.Tianjin 300384,China;College of Basic Science,Tianjin Agricultural University,Tianjin 300384,China;College of Fisheries Science,Tianjin Agricultural University,Tianjin 300384,China)
出处
《环境科学与技术》
CAS
CSCD
北大核心
2022年第8期100-106,共7页
Environmental Science & Technology
基金
国家自然科学基金面上项目(32172978,31772857)
甘肃省科技计划项目民生科技专项(21CX6NP223)
天津市教委科技计划项目(2020ZD06)
中央引导地方科技发展专项(21ZYCGSN00500)。
关键词
SVM回归
产毒微囊藻基因型
微囊藻毒素
相关系数
预测模型
support vector machine regression
toxic Microcystis genotype
microcystin
correlation coefficient
prediction model
