Identifying and Analyzing Nonlinear Concentration-Response Relationships in Tox21 Thyroid Receptor Assays Open Access

Kong, Seonhee (Spring 2021)

Permanent URL:


Background: Environmental thyroid disrupting chemicals (TDCs) can exhibit nonlinear including nonmonotonic dose response (NMDR) behaviors at low doses. NMDR presents a serious challenge because it could be difficult to accurately predict the risk of exposure to a particular chemical. New risk assessment methodologies produce toxicity testing results that are efficient and suitable for human application via in vitro assays. This study was aimed to analyze the concentration-response curves and identify nonlinearities including NMDR in the Tox21 high-throughput screening (HTS) data of thyroid receptor (TR) assays.


Methods: The Tox21 quantitative HTS TR assay data from the National Center for Advancing Translational Sciences (NCATS) were used. For the TR assays conducted in the antagonist or agonist mode and the viability assay, 10,496 chemicals (10K) were screened each. Among averaged concentration-response curves, the curves which have either low variance levels or response levels bound in a certain range were screened out. Correlation-based algorithm was applied for clustering followed by supervised learning algorithm for curve classification into five shapes – U, Bell, monotonic increasing, monotonic decreasing, and flat. Additional filtering was done based on viability and interference assays. For those remaining monotonic increasing or decreasing curves, their Hill coefficients were assessed for degree of nonlinearities.


Results: After all the screenings, 1,353 monotonic decreasing curves, 263 monotonic increasing curves, 577 U curves, 88 Bell curves were identified in antagonist mode assays. 16 monotonic decreasing curves, 19 monotonic increasing curves, 1 U curve, 9 Bell curves were identified in agonist mode assays. In antagonist mode assays, among known TDCs, TBBPA, pentachlorophenol, triclosan, and BPA exhibit as monotonic decreasing shapes, some of which have Hill coefficients considerably greater than 1 representing high degree of nonlinearities. Methimazole, propylthiouracil were classified as U shape.


Conclusions: The Tox21 TR assays contain many concentration-response curves that are either nonmonotonic or highly nonlinear even though they are monotonic, suggesting nonlinear dose response of endocrine disruptors can arise at the gene transcription level. Identifying nonlinear concentration-response relationship and underlying mechanisms will help provide a scientific basis for improving safety assessment of chemical products.

Table of Contents

1. Introduction 1

1.1. Thyroid system in general 1

1.2. Disruption of the thyroid system by environmental chemicals 2

1.2.1. Thyroid Disruptors and Mechanisms 2

1.2.2. General Health Outcome of TDCs 2

1.2.3. Nonlinear Dose-Response of TDCs 3

1.2.4. Toxicity Testing of Thyroid Disruptors 4

1.2.5. Purpose of This Study 6

2. Methods 7

2.1. Tox21 TR assays 7

2.2. Methods for clustering and classification of dose-response curves 9

2.2.1. Quality Control 9

2.2.2. Unsupervised Learning Algorithm (Clustering) 12

2.2.3. Supervised Learning Algorithm (Curve Classification) 13

2.2.4. Additional Filtering Based on Viability and Interference Assays 14

2.2.5. Hill Function to Assess Nonlinearities in Monotonic Curves 15

3. Results 16

3.1. Antagonist Assay 16

3.1.1. Statistics of Concentration-Response Curves in Each Shape 16

3.1.2. Scatter Plots for Lowest Response vs. Highest Response of Monotonically Shaped Curves 17

3.1.3. Scatter Plots for Inflection Point vs. Magnitude in NMDR Curves 19

3.1.4. Representative Concentration-Response Curves for Each Category 20

3.1.5. Hill Coefficient for Monotonic Shapes 23

3.2. Agonist Assay Results 26

3.2.1. Statistics of Concentration-Response Curves in Each Shape 26

3.2.2. Scatter Plots for Lowest Response vs. Highest Response in Each Monotonic Shape 27

3.2.3. Scatter Plots for Inflection Point vs. Magnitude in NMDR curves 28

3.2.4. Representative Curves for Each Category 29

3.2.5. Hill Coefficient for Monotonic Shapes 31

3.3. Comparison between Agonist and Antagonist Assay Results 33

4. Discussion 35

4.1. Linking to the Literatures 35

4.2. Pros. and Cons. of the Study 36

4.3. In relationship to other TH related assay results. 37

5. Conclusion 39

References 40

About this Master's Thesis

Rights statement
  • Permission granted by the author to include this thesis or dissertation in this repository. All rights reserved by the author. Please contact the author for information regarding the reproduction and use of this thesis or dissertation.
  • English
Research Field
Committee Chair / Thesis Advisor
Last modified

Primary PDF

Supplemental Files