Comparing Machine Learning Models for Aromatase (P450 19A1).

Aromatase, or cytochrome P450 19A1, catalyzes the aromatization of androgens to estrogens within the body. Changes in the activity of this enzyme can produce hormonal imbalances that can be detrimental to sexual and skeletal development. Inhibition of this enzyme can occur with drugs and natural products as well as environmental chemicals. Therefore, predicting potential endocrine disruption via exogenous chemicals requires that aromatase inhibition be considered in addition to androgen and estrogen pathway interference. Bayesian machine learning methods can be used for prospective prediction from the molecular structure without the need for experimental data. Herein, the generation and evaluation of multiple machine learning models utilizing different sources of aromatase inhibition data are described. These models are applied to two test sets for external validation with molecules relevant to drug discovery from the public domain. In addition, the performance of multiple machine learning algorithms was evaluated by comparing internal five-fold cross-validation statistics of the training data. These methods to predict aromatase inhibition from molecular structure, when used in concert with estrogen and androgen machine learning models, allow for a more holistic assessment of endocrine-disrupting potential of chemicals with limited empirical data and enable the reduction of the use of hazardous substances.

Comparison of Machine Learning Models for the Androgen Receptor.

The androgen receptor (AR) is a target of interest for endocrine disruption research, as altered signaling can affect normal reproductive and neurological development for generations. In an effort to prioritize compounds with alternative methodologies, the U.S. Environmental Protection Agency (EPA) used in vitro data from 11 assays to construct models of AR agonist and antagonist signaling pathways. While these EPA ToxCast AR models require in vitro data to assign a bioactivity score, Bayesian machine learning methods can be used for prospective prediction from molecule structure alone. This approach was applied to multiple types of data corresponding to the EPA's AR signaling pathway with proprietary software, Assay Central. The training performance of all machine learning models, including six other algorithms, was evaluated by internal 5-fold cross-validation statistics. Bayesian machine learning models were also evaluated with external predictions of reference chemicals to compare prediction accuracies to published results from the EPA. The machine learning model group selected for further studies of endocrine disruption consisted of continuous AC50 data from the February 2019 release of ToxCast/Tox21. These efforts demonstrate how machine learning can be used to predict AR-mediated bioactivity and can also be applied to other targets of endocrine disruption.

Machine Learning Models for Estrogen Receptor Bioactivity and Endocrine Disruption Prediction.

The U.S. Environmental Protection Agency (EPA) periodically releases in vitro data across a variety of targets, including the estrogen receptor (ER). In 2015, the EPA used these data to construct mathematical models of ER agonist and antagonist pathways to prioritize chemicals for endocrine disruption testing. However, mathematical models require in vitro data prior to predicting estrogenic activity, but machine learning methods are capable of prospective prediction from the molecular structure alone. The current study describes the generation and evaluation of Bayesian machine learning models grouped by the EPA's ER agonist pathway model using multiple data types with proprietary software, Assay Central. External predictions with three test sets of in vitro and in vivo reference chemicals with agonist activity classifications were compared to previous mathematical model publications. Training data sets were subjected to additional machine learning algorithms and compared with rank normalized scores of internal five-fold cross-validation statistics. External predictions were found to be comparable or superior to previous studies published by the EPA. When assessing six additional algorithms for the training data sets, Assay Central performed similarly at a reduced computational cost. This study demonstrates that machine learning can prioritize chemicals for future in vitro and in vivo testing of ER agonism.

Bioaccessibility of metals in alloys: evaluation of three surrogate biofluids.

Bioaccessibility in vitro tests measure the solubility of materials in surrogate biofluids. However, the lack of uniform methods and the effects of variable test parameters on material solubility limit interpretation. One aim of this study was to measure and compare bioaccessibility of selected economically important alloys and metals in surrogate physiologically based biofluids representing oral, inhalation and dermal exposures. A second aim was to experimentally test different biofluid formulations and residence times in vitro. A third aim was evaluation of dissolution behavior of alloys with in vitro lung and dermal biofluid surrogates. This study evaluated the bioaccessibility of sixteen elements in six alloys and 3 elemental/metal powders. We found that the alloys/metals, the chemical properties of the surrogate fluid, and residence time all had major impacts on metal solubility. The large variability of bioaccessibility indicates the relevancy of assessing alloys as toxicologically distinct relative to individual metals.

Exploiting lipid-free tubing passive samplers and embryonic zebrafish to link site specific contaminant mixtures to biological responses.

The Biological Response Indicator Devices Gauging Environmental Stressors (BRIDGES) bio-analytical tool was developed in response to the need for a quantitative technology for assessing the toxicity of environmentally relevant contaminant mixtures. This tool combines passive samplers with the embryonic zebrafish model. When applied in an urban river it effectively linked site specific, bioavailable contaminant mixtures to multiple biological responses. Embryonic zebrafish exposed to extracts from lipid-free passive samplers that were deployed at five locations, within and outside of the Portland Harbor Superfund Megasite, displayed different responses. Six of the eighteen biological responses observed in 941 exposed zebrafish were significantly different between sites. This demonstrates the sensitivity of the bio-analytical tool for detecting spatially distinct toxicity in aquatic systems; bridging environmental exposure to biological response.

Selenium accumulation patterns in lotic and lentic aquatic systems.

Selenium (Se) concentrations in water column, sediment and insect compartments were measured over 3 years, in conjunction with selected physicochemical parameters, from lotic (flowing water) and lentic (standing water) sites within a single watershed in Utah, USA. There was evidence for steady-state concentrations of total [Se] in the insects, sediment and detritus, while there was no correlation between these concentrations and the concentration in surface water. Insect Se burden may therefore provide a more accurate measurement of food web accumulation risk than surface water Se concentration. The importance of organism-specific factors on Se transfer to higher trophic levels was revealed by the steady-state Se body burden within the same insect taxa occupying similar environmental compartments in both aquatic systems. Additionally, however, insect Se body burdens, even within similar taxa, were up to 7 times greater within the lentic compared with the lotic system, and site-specific biogeochemical processes are also likely to play a role in the pattern and level of Se accumulation between hydrogeochemically different aquatic systems occurring within the same watershed. Though a site-specific relationship was apparent between organic content and sediment and detritus Se concentrations, this factor did not account for insect Se accumulation differences between the lotic and lentic aquatic habitats. If carbon content is to be used as a site-specific predictor of Se accumulation potential, further investigations of it's influence on the food web accumulation rate of Se are required.