• Bulletin of the Korean Chemical Society
• /
• 제33권2호
• /
• pp.613-619
• /
• 2012

We applied several machine learning methods for developing QSAR models for prediction of acute toxicity to fathead minnow. The multiple linear regression (MLR) and artificial neural network (ANN) method were applied to predict 96 h $LC_{50}$ (median lethal concentration) of 555 chemical compounds. Molecular descriptors based on 2D chemical structure were calculated by PreADMET program. The recursive partitioning (RP) model was used for grouping of mode of actions as reactive or narcosis, followed by MLR method of chemicals within the same mode of action. The MLR, ANN, and two RP-MLR models possessed correlation coefficients ($R^2$) as 0.553, 0.618, 0.632, and 0.605 on test set, respectively. The consensus model of ANN and two RP-MLR models was used as the best model on training set and showed good predictivity ($R^2$=0.663) on the test set.

• 한국환경독성학회:학술대회논문집
• /
• 한국환경독성학회 2001년도 추계심포지움 및 학술발표회:환경오염의 생체지표를 이용한 위해성 평가의 최근 기법
• /
• pp.89-89
• /
• 2001

• 대한환경공학회지
• /
• 제27권1호
• /
• pp.67-74
• /
• 2005

기존의 방류수 모니터링에서는 개별 오염물질들의 농도를 기준으로 독성을 평가하였다. 그러나 많은 연구자들에 의해서 오염물질들이 공존하는 상황에서 나타나는 독성은 그들 간의 상호작용을 통해서 혼합독성의 형태로 나타난다고 보고되고 있다. 본 연구에서는 GC/MS 분석을 통해 방류수 중에 존재하는 주요 독성 기여 오염 물질들을 분석하고, Independent Action(IA), Concentration Addition(CA), Effect Summation(ES) 모델을 사용하여 방류수의 혼합독성을 상호 비교 평가하였다. GC/MS로 분석된 오염물질을 대상으로 D. magna 기준 독성 평가를 실시하였고, 10가지의 주요 독성 기여 오염물질을 선별하였다. Chloroneb, butylbenzylphthalate, pendimethaline, di-n-butylphthalate, di-iso-butylphthalate, diazinon, isofenphos, 2-chlorophenol, 2,4,6-trichlorophenol 과 p-octylphenol을 주요 오염물질로 선정하여 혼합독성 평가를 실시하였다. 혼합독성 평가 결과는 IA 예측모델과 매우 높은 상관성($r^2\;=\;0.8475$)을 나타내었다. ES와 CA 모델은 IA 모델과 비교하여 혼합독성 결과와 매우 낮은 상관성을 나타내었으며, 특히 ES는 실측값을 5배나 과도하게 예측하였다. 이러한 결과를 통해서 전남지역 방류수에 존재하는 주요 오염물질들의 혼합독성은 IA 모델을 통해 예측이 가능할 것으로 판단된다.

• Environmental Analysis Health and Toxicology
• /
• 제30권sup호
• /
• pp.7.1-7.10
• /
• 2015

Objectives For successful adoption of legislation controlling registration and assessment of chemical substances, it is important to obtain sufficient toxicological experimental evidence and other related information. It is also essential to obtain a sufficient number of predicted risk and toxicity results. Particularly, methods used in predicting toxicities of chemical substances during acquisition of required data, ultimately become an economic method for future dealings with new substances. Although the need for such methods is gradually increasing, the-required information about reliability and applicability range has not been systematically provided. Methods There are various representative environmental and human toxicity models based on quantitative structure-activity relationships (QSAR). Here, we secured the 10 representative QSAR-based prediction models and its information that can make predictions about substances that are expected to be regulated. We used models that predict and confirm usability of the information expected to be collected and submitted according to the legislation. After collecting and evaluating each predictive model and relevant data, we prepared methods quantifying the scientific validity and reliability, which are essential conditions for using predictive models. Results We calculated predicted values for the models. Furthermore, we deduced and compared adequacies of the models using the Alternative non-testing method assessed for Registration, Evaluation, Authorization, and Restriction of Chemicals Substances scoring system, and deduced the applicability domains for each model. Additionally, we calculated and compared inclusion rates of substances expected to be regulated, to confirm the applicability. Conclusions We evaluated and compared the data, adequacy, and applicability of our selected QSAR-based toxicity prediction models, and included them in a database. Based on this data, we aimed to construct a system that can be used with predicted toxicity results. Furthermore, by presenting the suitability of individual predicted results, we aimed to provide a foundation that could be used in actual assessments and regulations.

• 한국환경보건학회지
• /
• 제48권3호
• /
• pp.159-166
• /
• 2022

Background: A quantitative structure-activity relationship (QSAR) model was adopted in the Registration, Evaluation, Authorization, and Restriction of Chemicals (REACH, EU) regulations as well as the Act on Registration, Evaluation, etc. of Chemicals (AREC, Republic of Korea). It has been previously used in the registration of chemicals. Objectives: In this study, we investigated the correlation between the predicted data provided by three prediction programs using a QSAR model and actual experimental results (acute fish, daphnia magna toxicity). Through this approach, we aimed to effectively conjecture on the performance and determine the most applicable programs when designating toxic substances through the AREC. Methods: Chemicals that had been registered and evaluated in the Toxic Chemicals Control Act (TCCA, Republic of Korea) were selected for this study. Two prediction programs developed and operated by the U.S. EPA - the Ecological Structure-Activity Relationship (ECOSAR) and Toxicity Estimation Software Tool (T.E.S.T.) models - were utilized along with the TOPKAT (Toxicity Prediction by Komputer Assisted Technology) commercial program. The applicability of these three programs was evaluated according to three parameters: accuracy, sensitivity, and specificity. Results: The prediction analysis on fish and daphnia magna in the three programs showed that the TOPKAT program had better sensitivity than the others. Conclusions: Although the predictive performance of the TOPKAT program when using a single predictive program was found to perform well in toxic substance designation, using a single program involves many restrictions. It is necessary to validate the reliability of predictions by utilizing multiple methods when applying the prediction program to the regulation of chemicals.

• 한국환경보건학회지
• /
• 제49권2호
• /
• pp.89-98
• /
• 2023

Background: Organophosphorus flame retardants (OPFRs) are a group of chemical substances used in building materials and plastic products to suppress or mitigate the combustion of materials. Although OPFRs are generally used in mixed form, information on their mixture toxicity is quite scarce. Objectives: This study aims to elucidate the toxicity and determine the types of interaction (e.g., synergistic, additive, and antagonistic effect) of OPFRs mixtures. Methods: Nine organophosphorus flame retardants, including TEHP (tris(2-ethylhexyl) phosphate) and TDCPP (tris(1,3-dichloro-2-propyl) phosphate), were selected based on indoor dust measurement data in South Korea. Nine OPFRs were exposed to the luminescent bacteria Aliivibrio fischeri for 30 minutes and the human hepatocyte cell line HepG2 for 48 hours. Chemicals with significant toxicity were only used for mixture toxicity tests in HepG2. In addition, the observed EC_x values were compared with the predicted toxicity values in the CA (concentration addition) prediction model, and the MDR (model deviation ratio) was calculated to determine the type of interaction. Results: Only four chemicals showed significant toxicity in the luminescent bacteria assays. However, EC₅₀ values were derived for seven out of nine OPFRs in the HepG2 assays. In the HepG2 assays, the highest to lowest EC₅₀ were in the order of the molecular weight of the target chemicals. In the further mixture tests, most binary mixtures show additive interactions except for the two combinations that have TPhP (triphenyl phosphate), i.e., TPhP and TDCPP, and TPhP and TBOEP (tris(2-butoxyethyl) phosphate). Conclusions: Our data shows OPFR mixtures usually have additivity; however, more research is needed to find out the reason for the synergistic effect of TPhP. Also, the mixture experimental dataset can be used as a training and validation set for developing the mixture toxicity prediction model as a further step.

• Korean Chemical Engineering Research
• /
• 제61권4호
• /
• pp.523-541
• /
• 2023

3D 프린터의 활용이 높아짐에 따라 발생하는 화학물질에 대한 노출 빈도가 증가하고 있다. 그러나 3D 프린팅 발생 화학물질의 독성 및 유해성에 대한 연구는 미비하며, 분자 구조 데이터의 결측치로 인해 in silico 기법을 사용한 독성예측 연구는 저조한 실정이다. 본 연구에서는 화학물질의 분자구조 정보를 나타내는 주요 분자표현자의 결측치를 보간하여 3D 프린팅의 독성 및 유해성을 예측한 Data-centric QSAR 모델을 개발하였다. 먼저 MissForest 알고리즘을 사용해 3D 프린팅으로 발생되는 유해물질의 분자표현자 결측치를 보완하였으며, 서로 다른 4가지 기계학습 모델(결정트리, 랜덤포레스트, XGBoost, SVM)을 기반으로 Data-centric QSAR 모델을 개발하여 생물 농축 계수(Log BCF)와 옥탄올-공기분배계수(Log Koa), 분배계수(Log P)를 예측하였다. 또한, 설명 가능한 인공지능(XAI) 방법론 중 TreeSHAP (SHapley Additive exPlanations) 기법을 활용하여 Data-centric QSAR 모델의 신뢰성을 입증하였다. MissForest 알고리즘 기반 결측지 보간 기법은, 기존 분자구조 데이터에 비하여 약 2.5배 많은 분자구조 데이터를 확보할 수 있었다. 이를 바탕으로 개발된 Data-centric QSAR 모델의 성능은 Log BCF, Log Koa와 Log P를 각각 73%, 76%, 92% 의 예측 성능으로 예측할 수 있었다. 마지막으로 Tree-SHAP 분석결과 개발된 Data-centric QSAR 모델은 각 독성치와 물리적으로 상관성이 높은 분자표현자를 통하여 선택함을 설명할 수 있었고 독성 정보에 대한 높은 예측 성능을 확보할 수 있었다. 본 연구에서 개발한 방법론은 다른 프린팅 소재나 화학공정, 그리고 반도체/디스플레이 공정에서 발생 가능한 오염물질의 독성 및 인체 위해성 평가에 활용될 수 있을 것으로 사료된다.

• 전자통신동향분석
• /
• 제38권1호
• /
• pp.46-54
• /
• 2023

The announcement of the US Environmental Protection Agency that it will stop conducting or funding experimental studies on mammals by 2035 should prioritize ongoing efforts to develop and use alternative toxicity screening methods to animal testing. Toxicity screening is likely to be further developed considering the combination of human-induced pluripotent-stem-cell-derived organ-on-a-chip and multielectrode array (MEA) technologies. We briefly review the current status of MEA technology and MEA-based neuropharmacological drug screening using various cellular model systems. Highlighting the coronavirus disease pandemic, we shortly comment on the importance of early prediction of toxicity by applying artificial intelligence to the development of rapid screening methods.

• 응용통계연구
• /
• 제37권3호
• /
• pp.355-380
• /
• 2024

본 연구에서는 OECD test guideline 데이터를 이용하여 graph기반 딥러닝 모델들의 성능을 비교하고자 한다. OECD TG는 화학물질들이 인체와 환경에 미칠 잠재적 영향에 대해 시험하는 방법이며, 많은 실험이 동물실험을 통해 독성을 확인한다. 동물실험은 많은 시간과 비용이 들며, 윤리적 이슈가 있어 대안을 찾거나 최소화하는 방법들이 연구되고 있다. 딥러닝은 화학물질을 활용하는 다양한 분야에서 사용되고 있으며, 독성예측 분야에도 사용되고 있으며, 특히 graph 기반 모델에 대한 연구가 활발하다. 우리의 목표는 OECD TG 데이터에 대한 graph기반 딥러닝 모델들의 성능을 비교하여 가장 성능이 좋은 모델을 찾는 것이다. 우리는 OECD에서 운영하는 웹사이트 eChemportal.org에서 OECD TG를 따른 결과를 수집하였으며, 전처리 과정을 통해 학습이 불가능하거나 부적절한 화학물질은 제거하였다. 수집된 OECD TG데이터와 화학물질 특성 예측 성능의 벤치마크 데이터셋인 MoleculeNet 데이터를 활용하여 5개의 graph기반 모델들의 독성 예측 성능을 비교하였다.

• 한국환경독성학회:학술대회논문집
• /
• 한국환경독성학회 2001년도 추계심포지움 및 학술발표회:환경오염의 생체지표를 이용한 위해성 평가의 최근 기법
• /
• pp.91-91
• /
• 2001