環保專案成果報告查詢系統

專案計畫編號
經費年度	109
計畫經費	6,624 千元
預算科目	化學物質評估與管理-毒性及關注化學物質管理
專案開始日期	2020/11/05
專案結束日期	2021/11/30
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	國立成功大學
專案分類	化學物質管理
中文關鍵字	危害交叉參照、高通量毒性測試、化學資訊模組、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 incident, computational toxicology
協同主持人	侯文哲、陳容甄、張偉翔、許昺奇
共同主持人
計畫聯絡人	陳秀玲
計畫聯絡電話	06-2353535#5864
計畫聯絡信箱	hsiulinchen@mail.ncku.edu.tw

中文摘要	本計畫參考美國Tox21及ToxCast之研究成果，希望藉由體外毒性測試及計算模擬的狀況下，可瞭解未知化學物質及混合化學物質之體外毒性測試結果，並結合高通量篩選毒理資料庫與計算毒理學，完成毒性高通量風險預測平臺之前期規劃，以利後續既有化學物質、新化學物質與環境污染有關化學品之風險評估與風險管理加值應用。本年度已完成彙整民國94年至110年臺灣發生環境事故中所涉及化學物質清單、事故場所類型、災害類型及發生地區與土壤及地下水污染場址之相關資料，整理出210項化學物質清單。已完成彙整及研析經濟合作暨發展組織(Organisation for Economic Co-operation and Development, OECD)、美國、歐盟及日本之交叉參照指引，並建構本土化學物質交叉參照標準流程。交叉參照之應用係由土壤及地下水污染整治網資料中，地下水污染物較多之場址篩選出4個化學物質（1,1,2-三氯乙烷、1,1,2,2-四氯乙烷、甲苯及甲基苯乙烯）作為範例。1,1,2-三氯乙烷及甲苯之預測結果以MACCS指紋進行交叉參照所得到之預測值與實際值最相近，而1,1,2,2-四氯乙烷之半數致死劑量(lethal dose 50%, LD50)、未觀察到不良效應程度(no observable adverse effect level, NOAEL)與最低觀察到不良效應程度(lowest observed adverse effect level, LOAEL)毒性預測則分別以PubChem、Circular及Extended指紋所取得之相似物進行交叉參照後之結果與實際值較為相近。另彙整及探討OECD、聯合國及世界衛生組織之多重化學物質大量洩漏至環境之評估與管理資訊，及洩漏至環境之處理原則及方法，完成蒐集美國、歐盟、OECD及日本已公布之最新環境採樣方法及傳輸模式資訊；亦完成以一場次國內污染場址進行研析及辨識其環境混合物染污之成分及危害資訊。毒性測試模式參照國際公開之替代測試規範，建立4個細胞株，針對細胞死亡、3種肝細胞代謝酵素活性、粒線體功能損傷與氧化壓力、2種內分泌干擾作用、細胞週期、細胞凋亡及微核總共11項毒性測試終點，進行高通量(high throughput)及高含量(high content)測試模組之建置，並針對篩選出之化學物質與環境真實樣本進行相關毒性評估。模型建置測試結果顯示，肝臟酵素活性之模型需再進行校正，其餘模型皆能提供一定準確程度之資料進行後續分析。環境真實樣本經肝臟毒性相關之模型試驗後，顯示編號24、34、37、38、39、40以及41號樣本皆具有肝臟毒性。在大鼠口服LD50的定量結構活性關係 (quantitative structure activity relationship, QSAR) 模型預測結果中，毒性預測軟體工具 (Toxicity Estimation Software Tool, T.E.S.T.) 中的階層式集群 (Hierarchical clustering method) 演算法的毒性分級準確率較好，大約為66.7%；而生殖毒性的QSAR模型預測結果中，所使用的QSAR模型在外部驗證的專一性表現較差，約只有44%至67%，但所有模型一致性的表現都在7成以上，代表模型在整體預測上還是相當可信，因此，QSAR具有快速評估篩選環境混合污染物有機成分毒性的潛力。另外，將我國化學清單（172筆）與ToxCast資料（8574筆）串接並完成自組織映射網路 (Self-Organizing Maps, SOM) 模式建立，優先篩選出全部化學物質皆具有ToxCast資料的化學結構相似群組，進行RAx的模式建立及驗證。此外，對於毒性資料未知者，也篩選出化學結構相似群組，該群組有未知毒性之我國化學清單，其他物質則具有豐富的ToxCast資料，將此群組的化學資料特性應用於RAx的模式建立，可用以瞭解未來將RAx應用於未知化學物質的毒理資料推估狀況。毒理學優先化指數 (Toxicological Priority Index, ToxPi) 評估結果顯示，191項化學物質中，異丙醇、甲基第三丁基醚、1,1,2,2-四氯乙烷及異丙苯為16種評估情境中應優先關注之化學物質。而評估之參數資訊來源及參數權重設定將影響化學物質之優先關注排序，應依據關切議題設定參數權重，使評估結果更切合評估目的。
中文關鍵字	危害交叉參照、高通量毒性測試、化學資訊模組、化學物質災害、計算毒理

中文摘要

本計畫參考美國Tox21及ToxCast之研究成果，希望藉由體外毒性測試及計算模擬的狀況下，可瞭解未知化學物質及混合化學物質之體外毒性測試結果，並結合高通量篩選毒理資料庫與計算毒理學，完成毒性高通量風險預測平臺之前期規劃，以利後續既有化學物質、新化學物質與環境污染有關化學品之風險評估與風險管理加值應用。本年度已完成彙整民國94年至110年臺灣發生環境事故中所涉及化學物質清單、事故場所類型、災害類型及發生地區與土壤及地下水污染場址之相關資料，整理出210項化學物質清單。已完成彙整及研析經濟合作暨發展組織(Organisation for Economic Co-operation and Development, OECD)、美國、歐盟及日本之交叉參照指引，並建構本土化學物質交叉參照標準流程。交叉參照之應用係由土壤及地下水污染整治網資料中，地下水污染物較多之場址篩選出4個化學物質（1,1,2-三氯乙烷、1,1,2,2-四氯乙烷、甲苯及甲基苯乙烯）作為範例。1,1,2-三氯乙烷及甲苯之預測結果以MACCS指紋進行交叉參照所得到之預測值與實際值最相近，而1,1,2,2-四氯乙烷之半數致死劑量(lethal dose 50%, LD50)、未觀察到不良效應程度(no observable adverse effect level, NOAEL)與最低觀察到不良效應程度(lowest observed adverse effect level, LOAEL)毒性預測則分別以PubChem、Circular及Extended指紋所取得之相似物進行交叉參照後之結果與實際值較為相近。另彙整及探討OECD、聯合國及世界衛生組織之多重化學物質大量洩漏至環境之評估與管理資訊，及洩漏至環境之處理原則及方法，完成蒐集美國、歐盟、OECD及日本已公布之最新環境採樣方法及傳輸模式資訊；亦完成以一場次國內污染場址進行研析及辨識其環境混合物染污之成分及危害資訊。毒性測試模式參照國際公開之替代測試規範，建立4個細胞株，針對細胞死亡、3種肝細胞代謝酵素活性、粒線體功能損傷與氧化壓力、2種內分泌干擾作用、細胞週期、細胞凋亡及微核總共11項毒性測試終點，進行高通量(high throughput)及高含量(high content)測試模組之建置，並針對篩選出之化學物質與環境真實樣本進行相關毒性評估。模型建置測試結果顯示，肝臟酵素活性之模型需再進行校正，其餘模型皆能提供一定準確程度之資料進行後續分析。環境真實樣本經肝臟毒性相關之模型試驗後，顯示編號24、34、37、38、39、40以及41號樣本皆具有肝臟毒性。在大鼠口服LD50的定量結構活性關係 (quantitative structure activity relationship, QSAR) 模型預測結果中，毒性預測軟體工具 (Toxicity Estimation Software Tool, T.E.S.T.) 中的階層式集群 (Hierarchical clustering method) 演算法的毒性分級準確率較好，大約為66.7%；而生殖毒性的QSAR模型預測結果中，所使用的QSAR模型在外部驗證的專一性表現較差，約只有44%至67%，但所有模型一致性的表現都在7成以上，代表模型在整體預測上還是相當可信，因此，QSAR具有快速評估篩選環境混合污染物有機成分毒性的潛力。另外，將我國化學清單（172筆）與ToxCast資料（8574筆）串接並完成自組織映射網路 (Self-Organizing Maps, SOM) 模式建立，優先篩選出全部化學物質皆具有ToxCast資料的化學結構相似群組，進行RAx的模式建立及驗證。此外，對於毒性資料未知者，也篩選出化學結構相似群組，該群組有未知毒性之我國化學清單，其他物質則具有豐富的ToxCast資料，將此群組的化學資料特性應用於RAx的模式建立，可用以瞭解未來將RAx應用於未知化學物質的毒理資料推估狀況。毒理學優先化指數 (Toxicological Priority Index, ToxPi) 評估結果顯示，191項化學物質中，異丙醇、甲基第三丁基醚、1,1,2,2-四氯乙烷及異丙苯為16種評估情境中應優先關注之化學物質。而評估之參數資訊來源及參數權重設定將影響化學物質之優先關注排序，應依據關切議題設定參數權重，使評估結果更切合評估目的。

中文關鍵字

危害交叉參照、高通量毒性測試、化學資訊模組、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	強化我國化學物質毒性高通量風險預測平臺前期建置計畫.pdf	24MB	計畫成果報告	2022/1/3 上午 12:00:00

摘要	This report is based on the US Tox21 and ToxCast results to explore the in vitro toxicity testing results of the substance of unknown or variable composition, complex reaction products or biological materials (UVCBs) through in vitro toxicity testing or computational simulation. Also, by combining the high-throughput screening toxicology database and computational toxicology to complete the preliminary establishment of the high-throughput platform of chemical hazard assessment and risk prediction for the future usage on new and existing chemical substances and environmental pollutant related chemicals’ risk assessment and management. This report had compiled a chemical list of 210 chemicals along with related information from the Taiwan emergence chemical incidents and contaminated sites since 2005. The Read Across Guidelines (and databases used in the read-across application) proposed by the developed countries or international organizations were collected to build the Taiwan Chemical Substance Read Across Procedure. The read across application has been carried out with examples based on the Taiwan Chemical Substance Read Across Procedure. The read across prediction for 1,1,2-trichloroethane and toluene by MACCS fingerprint showed a closest result to its experimental results, while for 1,1,2,2-tetrachloroethane, the analogs obtained by PubChem, Circular and Extended fingerprints showed a better prediction on LD50, NOAEL and LOAEL. Furthermore, the information and variables required to assess and to manage the multiple chemicals spill to the environment were also compiled. The managing principles to handle the multiple chemical spill into the environment were explored and consolidated. Moreover, the collection of.latest chemical sampling method and transmission model were completed. An exploration and identification of complex chemical substances and its hazard from the local contaminated site has been done.The internationally published alternative test guidelines were used to build the relevant high-throughput screening (HTS) and high content screening (HCS) test modules using 4 cell lines and evaluating 11 toxicity endpoints including cell death, 3 liver cell metabolic enzyme activities, mitochondrial damage, oxidative stress, 2 endocrine disrupting effects, cell cycle arrest, apoptosis, and micronuclei assay. The selected chemical substances and samples acquiring from the environment will be applied to relevant test modules for toxicity assessment. The results of the test model building show that the liver enzyme activity model needs to be calibrated, and the rest of the models can provide a certain accuracy level for subsequent analysis. Samples from the environment have undergone liver toxicity-related model tests, and it has been shown that samples 24, 34, 37, 38, 39, 40, and 41 have liver toxicity. The hierarchical clustering method algorithm of the Toxicity Estimation Software Tool (T.E.S.T.) module in Quantitative Structure Activity Relationship (QSAR) prediction model showed a better result for the rat oral LD50 toxicity classification with an accuracy of 66.7%. The reproductive and developmental toxicity prediction in all of the QSAR models showed poor specificity performance for external verification, however, the concordance performance for all QSAR models are greater than 70%, which indicates that the overall prediction of the models is still convincing. Hence, QSAR could be an emerging tool for rapid assessing and screening the environmental complex contaminants and organic components. The chemical list mentioned in the earlier part (172 chemicals with CASRN) were merged with the ToxCast database chemical list (8574 chemicals with CASRN), and the merged chemical substances list was then used to build the self-organizing map (SOM) module.Then, the chemicals in bin 217 and bin 305 of SOM were priorly selected for the read-across procedure building and verification, in order to understand the read across application to predict the hazard and toxicity data for unknown chemicals. The results of 16 scenarios show that among 191 chemicals, isopropanol, methyl tert-butyl ether, 1,1,2,2-tetrachloroethane and cumene are the chemicals that get the high Toxicological Priority Index (ToxPi) score, should be given priority attention. Besides, the information source of the parameters and the setting of parameter weights will affect the priority ranking of chemicals. So that the parameter weights should be set according to the issues of concern will make the evaluation results more suitable for the evaluation purpose.
英文關鍵字	Read-across (RAx), high throughput toxicity testing, cheminformatics, chemical substances incident, computational toxicology

摘要

This report is based on the US Tox21 and ToxCast results to explore the in vitro toxicity testing results of the substance of unknown or variable composition, complex reaction products or biological materials (UVCBs) through in vitro toxicity testing or computational simulation. Also, by combining the high-throughput screening toxicology database and computational toxicology to complete the preliminary establishment of the high-throughput platform of chemical hazard assessment and risk prediction for the future usage on new and existing chemical substances and environmental pollutant related chemicals’ risk assessment and management. This report had compiled a chemical list of 210 chemicals along with related information from the Taiwan emergence chemical incidents and contaminated sites since 2005. The Read Across Guidelines (and databases used in the read-across application) proposed by the developed countries or international organizations were collected to build the Taiwan Chemical Substance Read Across Procedure. The read across application has been carried out with examples based on the Taiwan Chemical Substance Read Across Procedure. The read across prediction for 1,1,2-trichloroethane and toluene by MACCS fingerprint showed a closest result to its experimental results, while for 1,1,2,2-tetrachloroethane, the analogs obtained by PubChem, Circular and Extended fingerprints showed a better prediction on LD50, NOAEL and LOAEL. Furthermore, the information and variables required to assess and to manage the multiple chemicals spill to the environment were also compiled. The managing principles to handle the multiple chemical spill into the environment were explored and consolidated. Moreover, the collection of.latest chemical sampling method and transmission model were completed. An exploration and identification of complex chemical substances and its hazard from the local contaminated site has been done.The internationally published alternative test guidelines were used to build the relevant high-throughput screening (HTS) and high content screening (HCS) test modules using 4 cell lines and evaluating 11 toxicity endpoints including cell death, 3 liver cell metabolic enzyme activities, mitochondrial damage, oxidative stress, 2 endocrine disrupting effects, cell cycle arrest, apoptosis, and micronuclei assay. The selected chemical substances and samples acquiring from the environment will be applied to relevant test modules for toxicity assessment. The results of the test model building show that the liver enzyme activity model needs to be calibrated, and the rest of the models can provide a certain accuracy level for subsequent analysis. Samples from the environment have undergone liver toxicity-related model tests, and it has been shown that samples 24, 34, 37, 38, 39, 40, and 41 have liver toxicity. The hierarchical clustering method algorithm of the Toxicity Estimation Software Tool (T.E.S.T.) module in Quantitative Structure Activity Relationship (QSAR) prediction model showed a better result for the rat oral LD50 toxicity classification with an accuracy of 66.7%. The reproductive and developmental toxicity prediction in all of the QSAR models showed poor specificity performance for external verification, however, the concordance performance for all QSAR models are greater than 70%, which indicates that the overall prediction of the models is still convincing. Hence, QSAR could be an emerging tool for rapid assessing and screening the environmental complex contaminants and organic components. The chemical list mentioned in the earlier part (172 chemicals with CASRN) were merged with the ToxCast database chemical list (8574 chemicals with CASRN), and the merged chemical substances list was then used to build the self-organizing map (SOM) module.Then, the chemicals in bin 217 and bin 305 of SOM were priorly selected for the read-across procedure building and verification, in order to understand the read across application to predict the hazard and toxicity data for unknown chemicals. The results of 16 scenarios show that among 191 chemicals, isopropanol, methyl tert-butyl ether, 1,1,2,2-tetrachloroethane and cumene are the chemicals that get the high Toxicological Priority Index (ToxPi) score, should be given priority attention. Besides, the information source of the parameters and the setting of parameter weights will affect the priority ranking of chemicals. So that the parameter weights should be set according to the issues of concern will make the evaluation results more suitable for the evaluation purpose.

英文關鍵字

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