(62d) Nonlinear Modeling of the Estrogen Receptor Activity of Environmental Chemicals Using Classification Algorithms and Single-Cell Level Data

Environmental emergencies such as floods and hurricanes increase the potential risk for contact, inhalation, and ingestion of the released toxic household, medical, and industrial chemicals [1]. Environmental chemicals, both synthetic and natural, have the potential to adversely affect the endocrine system posing long-term risks to human and wildlife reproductive and metabolic health by binding to the endocrine hormone receptors, like estrogen (ER) and androgen (AR) receptors [2]. Modeling the estrogenic activity of benchmark endocrine-disrupting chemicals will allow us to predict agonistic and antagonistic behavior of novel chemicals and complex mixtures and facilitate the decision-making process for mitigative actions.

Recent advancements in big-data analytics and machine learning algorithms aid the computational characterization of estrogenic activity of chemicals using cell-based high throughput assays [3-4]. In this work, high-throughput microscopy and single-cell image analysis followed by machine learning methods were used to construct a classification model of estrogen receptor activity by environmental chemicals based on the single-cell level data without any data reduction strategy. We used 55 different compounds in the experiments which were defined as agonists, antagonists, or inactive chemicals of the ER by the US Environmental Protection Agency (EPA) [5]. After preprocessing the data obtained from 70 size, shape, and intensity features from cell and nuclear compartments after treatment with each chemical, supervised machine learning analyses – Support Vector Machine (SVM), Random Forest (RF), Neural Network (NN) classifiers were trained to ultimately predict the estrogen receptor-α agonists or antagonists. The results show that the two-class non-linear SVM model predicts the outcome of the test data with more than 92% accuracy.

References

[1] Cooper, C.M. and Wardropper, C.B., 2021. Environmental, public health, and economic development perspectives at a Superfund site: AQ methodology approach. Journal of Environmental Management, 279, p.111571.

[2] EPA, 2022. What is Endocrine Disruption? https://www.epa.gov/endocrine-disruption/what-endocrine-disruption (Accessed on April 5, 2022).

[3] Ganesh, H.S., Beykal, B., Szafran, A.T., Stossi, F., Zhou, L., Mancini, M.A. and Pistikopoulos, E.N., 2021. Predicting the Estrogen Receptor Activity of Environmental Chemicals by Single-Cell Image Analysis and Data-driven Modeling. In Computer Aided Chemical Engineering (Vol. 50, pp. 481-486). Elsevier.

[4] Mukherjee, R., Beykal, B., Szafran, A.T., Onel, M., Stossi, F., Mancini, M.G., Lloyd, D., Wright, F.A., Zhou, L., Mancini, M.A. and Pistikopoulos, E.N., 2020. Classification of estrogenic compounds by coupling high content analysis and machine learning algorithms. PLoS computational biology, 16(9), p.e1008191.

[5] Judson, R.S., Magpantay, F.M., Chickarmane, V., Haskell, C., Tania, N., Taylor, J., Xia, M., Huang, R., Rotroff, D.M., Filer, D.L. and Houck, K.A., 2015. Integrated model of chemical perturbations of a biological pathway using 18 in vitro high-throughput screening assays for the estrogen receptor. Toxicological Sciences, 148(1), pp.137-154.