东南大学学报(自然科学版)
東南大學學報(自然科學版)
동남대학학보(자연과학판)
JOURNAL OF SOUTHEAST UNIVERSITY
2013年
6期
1274-1279
,共6页
袁金涛%浦跃朴%尹立红
袁金濤%浦躍樸%尹立紅
원금도%포약박%윤립홍
芳香胺%致癌性预测%线性判别%Ames%构效关系
芳香胺%緻癌性預測%線性判彆%Ames%構效關繫
방향알%치암성예측%선성판별%Ames%구효관계
aromatic amine%carcinogenicity prediction%linear discriminant analysis%Ames%quanti-tative structure-activity relationship
为探索 Ames 试验结果对预测芳香胺致癌活性的作用,采用芳香胺结构描述符或描述符与 Ames 结果相结合的方法建立了多个线性判别模型,对113个芳香胺的致癌性进行预测,并对各模型预测性能进行了比较.描述符与 Ames 结果相结合的方法分2种进行,一种是将 Ames 结果作为模型的描述符,另一种是先根据 Ames 结果将化合物分为两类,再建立模型.通过比较发现,基于2种方法建立的模型对113个芳香胺的致癌性预测正确率均为80.5%,略高于仅用结构描述符得到的模型的预测准确率77.0%,但预测正确率差异不具统计学意义.仅采用结构描述符的构效关系模型具有较满意结果;将 Ames 结果与 QSAR 模型相结合并不能显著提高芳香胺致癌性预测模型的预测性能.
為探索 Ames 試驗結果對預測芳香胺緻癌活性的作用,採用芳香胺結構描述符或描述符與 Ames 結果相結閤的方法建立瞭多箇線性判彆模型,對113箇芳香胺的緻癌性進行預測,併對各模型預測性能進行瞭比較.描述符與 Ames 結果相結閤的方法分2種進行,一種是將 Ames 結果作為模型的描述符,另一種是先根據 Ames 結果將化閤物分為兩類,再建立模型.通過比較髮現,基于2種方法建立的模型對113箇芳香胺的緻癌性預測正確率均為80.5%,略高于僅用結構描述符得到的模型的預測準確率77.0%,但預測正確率差異不具統計學意義.僅採用結構描述符的構效關繫模型具有較滿意結果;將 Ames 結果與 QSAR 模型相結閤併不能顯著提高芳香胺緻癌性預測模型的預測性能.
위탐색 Ames 시험결과대예측방향알치암활성적작용,채용방향알결구묘술부혹묘술부여 Ames 결과상결합적방법건립료다개선성판별모형,대113개방향알적치암성진행예측,병대각모형예측성능진행료비교.묘술부여 Ames 결과상결합적방법분2충진행,일충시장 Ames 결과작위모형적묘술부,령일충시선근거 Ames 결과장화합물분위량류,재건립모형.통과비교발현,기우2충방법건립적모형대113개방향알적치암성예측정학솔균위80.5%,략고우부용결구묘술부득도적모형적예측준학솔77.0%,단예측정학솔차이불구통계학의의.부채용결구묘술부적구효관계모형구유교만의결과;장 Ames 결과여 QSAR 모형상결합병불능현저제고방향알치암성예측모형적예측성능.
To explore the role of Ames test results in carcinogenicity prediction of aromatic amines, the carcinogenicity of 113 aromatic amines were predicted by several linear discriminant analysis models with structure descriptors or Ames results coupled with structure descriptors.And the predic-tive abilities of these different models were compared.The methods of combing Ames results with structure descriptors were carried out in two ways.One was that Ames results performed as descrip-tors of models;the other was that models were constructed after compounds were classified into two according to the Ames results.The comparison results show that the accuracies of the two models were both 80.5%,slightly higher than that of the model obtained only with structure descriptors (77.0%).But the accuracy difference of these methods is not statistically significant.The quantita-tive structure-activity relationship (QSAR)model obtained only with structure descriptors exhibits satisfactory results.Combining Ames results with the QSAR model cannot significantly improve the prediction ability of the carcinogenicity prediction model of aromatic amine.