環保專案成果報告查詢系統

專案計畫編號
經費年度	111
計畫經費	3,823 千元
預算科目	化學物質評估與管理-化學物質登錄審查
專案開始日期	2022/02/11
專案結束日期	2022/11/30
專案主持人	陳秀玲
主辦單位	化學局
承辦人	崔君至
執行單位	成功大學
專案分類	化學物質管理
中文關鍵字	交叉參照、高通量毒性測試、化學資訊學、化學物質災害、計算毒理
英文關鍵字	Read-across (RAx), high throughput toxicity testing, cheminformatics, chemical substances hazard, computational toxicology
協同主持人	侯文哲、陳容甄、張偉翔、薛欣達
共同主持人
計畫聯絡人	王琬茹
計畫聯絡電話	06-2353535#5870
計畫聯絡信箱	s78084024@gs.ncku.edu.tw

中文摘要	隨著工業與科技迅速發展，促使化學物質之應用範圍不同於以往，且新興化學物質數量持續增長。目前已登錄之化學物質數目中僅有少數的化學物質已知毒性，如何補足二者間差距即為現今化學物質使用與管理之重要議題。預測毒理學及計算毒理學的運用為國際間新興化學物質毒性測試提供了較節省時間及經費之方法，且藉由演算法亦可針對已存毒性資訊之化學物質進行權重篩選，進而有效地推動後續管理規劃。本年度包含三大工項：（一）強化本土多重化學物質危害交叉參照(Read-Across)資訊；（二）建立我國化學物質高通量(High throughput)毒性篩選模組；（三）應用化學資訊學(Cheminformatics)模組於高通量風險評估。已完成交叉參照(Read-across, RAx)結果預測與驗證，共預測8種化學物質與6種毒性，其總毒性(total toxicity)之接收者操作特徵曲線(Receiver operating characteristic curve, ROC曲線)下面積(AUC)皆高於0.8，且以線性回歸建立預測值與真實值之結果判定係數(Coefficient of determination, R2)高於0.7，表示總毒性預測結果落於可接受範圍，而單一毒性之預測結果ROC曲線下面積僅落於0.6-0.7左右，且部分毒性結果未達顯著。另完成多種化學物質之資料庫整合，並建立多重化學物質交叉參照資料篩選介面，可提供動物實驗與細胞實驗之平均值、最大值及最小值，以應用於RAx與毒理學優先指數(Toxicological Priority Index, ToxPi)模組之敏感度分析。建立腎上腺皮質癌細胞(H295R細胞)及人類腎臟近端小管上皮細胞(HK-2細胞)之測試模式，並以前期計畫與本年度所建立之高通量細胞模組應用於真實樣本之毒性探討，結果發現真實樣本中，可能具有內分泌干擾及肝臟毒性之物質。本年度自建2種水生急毒性定量構效關係(Quantitative Structure-Activity Relationship, QSAR)模型之R2分別為0.94及0.93，r test 2（測試集之實驗值與預測值相關係數）分別為0.70與0.71、Q2（10折交叉驗證相關係數）為0.63與0.55，皆在可接受範圍內，並將此2模型用於環境真實樣本之環境毒理預測，作為ToxPi參數。ToxPi模組建立與敏感度測試結果指出，五氯酚、2,4,6-三氯酚、甲醛、2,4,5-三氯酚及3,3’-二氯聯苯胺其ToxPi Score多為排名前5名，屬應優先關注之化學物質；真實環境樣本之採樣點優先關注排序結果亦顯示，石化廠址12個地下水水樣共有4個採樣點應優先納入管理考量。
中文關鍵字	交叉參照、高通量毒性測試、化學資訊學、化學物質災害、計算毒理

中文摘要

隨著工業與科技迅速發展，促使化學物質之應用範圍不同於以往，且新興化學物質數量持續增長。目前已登錄之化學物質數目中僅有少數的化學物質已知毒性，如何補足二者間差距即為現今化學物質使用與管理之重要議題。預測毒理學及計算毒理學的運用為國際間新興化學物質毒性測試提供了較節省時間及經費之方法，且藉由演算法亦可針對已存毒性資訊之化學物質進行權重篩選，進而有效地推動後續管理規劃。本年度包含三大工項：（一）強化本土多重化學物質危害交叉參照(Read-Across)資訊；（二）建立我國化學物質高通量(High throughput)毒性篩選模組；（三）應用化學資訊學(Cheminformatics)模組於高通量風險評估。已完成交叉參照(Read-across, RAx)結果預測與驗證，共預測8種化學物質與6種毒性，其總毒性(total toxicity)之接收者操作特徵曲線(Receiver operating characteristic curve, ROC曲線)下面積(AUC)皆高於0.8，且以線性回歸建立預測值與真實值之結果判定係數(Coefficient of determination, R2)高於0.7，表示總毒性預測結果落於可接受範圍，而單一毒性之預測結果ROC曲線下面積僅落於0.6-0.7左右，且部分毒性結果未達顯著。另完成多種化學物質之資料庫整合，並建立多重化學物質交叉參照資料篩選介面，可提供動物實驗與細胞實驗之平均值、最大值及最小值，以應用於RAx與毒理學優先指數(Toxicological Priority Index, ToxPi)模組之敏感度分析。建立腎上腺皮質癌細胞(H295R細胞)及人類腎臟近端小管上皮細胞(HK-2細胞)之測試模式，並以前期計畫與本年度所建立之高通量細胞模組應用於真實樣本之毒性探討，結果發現真實樣本中，可能具有內分泌干擾及肝臟毒性之物質。本年度自建2種水生急毒性定量構效關係(Quantitative Structure-Activity Relationship, QSAR)模型之R2分別為0.94及0.93，r test 2（測試集之實驗值與預測值相關係數）分別為0.70與0.71、Q2（10折交叉驗證相關係數）為0.63與0.55，皆在可接受範圍內，並將此2模型用於環境真實樣本之環境毒理預測，作為ToxPi參數。ToxPi模組建立與敏感度測試結果指出，五氯酚、2,4,6-三氯酚、甲醛、2,4,5-三氯酚及3,3’-二氯聯苯胺其ToxPi Score多為排名前5名，屬應優先關注之化學物質；真實環境樣本之採樣點優先關注排序結果亦顯示，石化廠址12個地下水水樣共有4個採樣點應優先納入管理考量。

中文關鍵字

交叉參照、高通量毒性測試、化學資訊學、化學物質災害、計算毒理

類型	檔名	檔案大小	說明	最新版公開日期
期末報告	多元化學物質高通量風險管理計畫.pdf	16MB		2023/1/10 上午 12:00:00

摘要	The number of emerging chemical substances grow faster and show their variety with the rapid development of industries and technologies. There are only 7.5% of the registered chemical substances have known toxicity. How to fill the data gap of the chemical substance with unknown toxicity become a critical issue in the management of chemical substances. Therefore, the application of computational toxicology, which is the integrative approach to toxicological research and chemical safety assessments via predictive modeling, provides a time-saving and cost-effective method for the toxicity testing of emerging chemical substances. Moreover, the approach may enhance the risk assessment and management of emerging chemical substances in the future. This report includes three major parts, which shows as follows, (1) Strengthening the hazard cross-reference (Read-Across) information of chemical substances, (2) Establishing a high-throughput toxicity screening model for chemical substances, (3) Applied chemistry Informatics (Cheminformatics) module for high-throughput risk assessment. We have built the toxicity prediction by Read Across and provided the validation methods in this report. For Projects 1, at least eight chemical substances were used to conduct the total toxicity prediction and different type toxicity prediction. The validation results suggested the areas under ROC curve (AUC) all over 0.8. However, the validation results of different type of toxicity were weak. Furthermore, we have established a platform for the integration of the databases, which could provide the average, maximum and minimum values of animal experiments and cell experiments from different databases. The dataset will use to build the modules of Read-across and ToxPi. For project 2, we have completed the establishment of high throughput cell models of H295R cells and HK-2 human renal proximal tubular epithelial cells, and applied the cell models established last year and this year to real water samples for high-throughput toxicological mechanism studies. The endocrine disruptors and liver toxicity were found in the real water samples obtained from groundwater of petroleum industries. In the future, the application of high-throughput toxicological study of environmental samples should add the high-throughput analysis results of chemical substances. For the QSAR model, the R2 (coefficient of determination) is 0.94 and 0.93, rtest2 (the correlation coefficient between the experimental value and the predicted value of the test set) is 0.7 and 0.71, and Q2 (10-fold cross-validation correlation coefficient) is 0.63 and 0.55, within the acceptable range for environmental toxicities. In addition, ToxPi module establishment and sensitivity test were completed. 6-Trichlorophenol, formaldehyde, 2,4,5-trichlorophenol and 3,3'-dichlorobenzidine all were ranked in the top 5 in ToxPi Score, and these substances should be paid priority attention. For real water samples, the results of the priority ranking of points suggested that 4 sampling sites should be prioritized into management considerations from 12 sampling areas.
英文關鍵字	Read-across (RAx), high throughput toxicity testing, cheminformatics, chemical substances hazard, computational toxicology

摘要

The number of emerging chemical substances grow faster and show their variety with the rapid development of industries and technologies. There are only 7.5% of the registered chemical substances have known toxicity. How to fill the data gap of the chemical substance with unknown toxicity become a critical issue in the management of chemical substances. Therefore, the application of computational toxicology, which is the integrative approach to toxicological research and chemical safety assessments via predictive modeling, provides a time-saving and cost-effective method for the toxicity testing of emerging chemical substances. Moreover, the approach may enhance the risk assessment and management of emerging chemical substances in the future. This report includes three major parts, which shows as follows, (1) Strengthening the hazard cross-reference (Read-Across) information of chemical substances, (2) Establishing a high-throughput toxicity screening model for chemical substances, (3) Applied chemistry Informatics (Cheminformatics) module for high-throughput risk assessment. We have built the toxicity prediction by Read Across and provided the validation methods in this report. For Projects 1, at least eight chemical substances were used to conduct the total toxicity prediction and different type toxicity prediction. The validation results suggested the areas under ROC curve (AUC) all over 0.8. However, the validation results of different type of toxicity were weak. Furthermore, we have established a platform for the integration of the databases, which could provide the average, maximum and minimum values of animal experiments and cell experiments from different databases. The dataset will use to build the modules of Read-across and ToxPi. For project 2, we have completed the establishment of high throughput cell models of H295R cells and HK-2 human renal proximal tubular epithelial cells, and applied the cell models established last year and this year to real water samples for high-throughput toxicological mechanism studies. The endocrine disruptors and liver toxicity were found in the real water samples obtained from groundwater of petroleum industries. In the future, the application of high-throughput toxicological study of environmental samples should add the high-throughput analysis results of chemical substances. For the QSAR model, the R2 (coefficient of determination) is 0.94 and 0.93, rtest2 (the correlation coefficient between the experimental value and the predicted value of the test set) is 0.7 and 0.71, and Q2 (10-fold cross-validation correlation coefficient) is 0.63 and 0.55, within the acceptable range for environmental toxicities. In addition, ToxPi module establishment and sensitivity test were completed. 6-Trichlorophenol, formaldehyde, 2,4,5-trichlorophenol and 3,3'-dichlorobenzidine all were ranked in the top 5 in ToxPi Score, and these substances should be paid priority attention. For real water samples, the results of the priority ranking of points suggested that 4 sampling sites should be prioritized into management considerations from 12 sampling areas.

英文關鍵字

Read-across (RAx), high throughput toxicity testing, cheminformatics, chemical substances hazard, computational toxicology