Welcome back. And now we'll be talking about the estrogen receptor bioactivity model, which is the most developed and the first example of an alternative being used in the endocrine receptor screening program to identify endocrine activity of chemicals. The model's been published in a variety of scientific publications, namely in the SOT's Toxicological Sciences journal. Where the integrated model of chemical perturbations of a biological pathway using 18 in vitro high-throughput screening assays for the estrogen receptor is published in Environmental Health Perspectives for a curated database of rodent uterotrophic bioactivity is published. And in the Journal, Environmental Science and Technology article screening chemicals for estrogen receptor bioactivity using a computational model is published. This provides a large part of the scientific basis for the estrogen receptor model as an alternative in the endocrine receptor screening program. So the estrogen receptor bioactivity model is a way to integrate data from 18 different high-throughput screening assays, in a way that replicates the biological pathway and indicates two different types of estrogen receptor activity, agonist activity or antagonist activity, or inhibitory or negative activity. That's indicated in this diagram. The assays are labeled as A1 through A18 in this diagram, with appropriate symbol as shown in the key. And then the R1 channel of agonist pathways indicated by the blue filled portions of the diagram. Then the antagonist pathway, our two channel of output from the model is indicated by the red filled portions of the diagram. You can see table indicating all of the different assays. These are from EPA's ToxCast research program in their High Throughput Screening Estrogen Receptor Assays, ranging across a variety of different targets in the biological pathway from the receptor to co-factors to downstream effects of receptor activation or inhibition. In a variety of different sources of high-throughput screening data in terms of companies and technologies. And this is really important, this range of a diversity of assay technologies that leads to the greater robustness of the collection of data and the model integrating those data. So you can see that, as indicated by the diagram, that there's outputs or activity in a collection of different assays that feeds into either the agonist pathway, so A1 through A16. So 16 assays all potentially contribute towards agonist activity, or the agonist pathway. And a subset of those, 13 of the 18, contribute toward the R2 or antagonist pathway. This next slide indicates output for an example chemical. I think in this case, it's bisphenol A, and it indicates the results of the 18 different individual assays in the left-hand panel. A concentration response is indicated here. So on the bottom is listed a chemical concentration in molar units, and the y-axis is the efficacy, the percent relative to estradiol, the natural ligand for estrogen receptor. So there's a relative zero to one scale of percent estrogen activity or estrogenic activity being indicated here. And you can see that for the 16 relevant agonist assays, A1 through A16, there's a concentration response indicated for each of the 16 assays. But they're all slightly different in terms of their potency, the concentration of which you start to see a rise in the concentration response curve. And there's also some variation in terms of their efficacy, the percent activity relevant to estradiol. All 16 of these data are integrated by the pathway model, in the right hand panel, into an R1 agonist response indicated by the blue concentration response curve. And what that indicates is the probability or agonist activity, and you can see that rising in higher concentration, integrating the results from those 16 relevant assays in the agonist channel, the R1 channel. There's little to no response for antagonist for this particular compound. It's a true agonist, so we wouldn't expect to see that. And then you see R3 through R12, which are assay technology or assay type specific. Pseudoreceptors or noise channels is one way to think of them. False signal is another way to think of those. And you can see, for example, the R3 concentration response curve, the R3 curve coming up initially. That’s because this A1 through A2, A3 binding assays are slightly more sensitive than the other assays. Go to the left-hand panel. See how they rise at a lower concentration, the first to come up, if you will, those black concentration response curves. So going back to the right-hand panel, you see a slight rise in the R3 channel output indicating the potential essay technology specific pulse signal. But as the other assays kick in, it recognizes that this is the model recognizes that this is in fact a true agonist signal. Multiple assays and multiple assay technologies are responding. And you see a concordant strong rise in the R1 signal, the agonist signal for this chemical, bisphenol A. This indicates how the pathway model can integrate results across the 18 different high-throughput screening assays to predict either agonism, antagonism, or indicate false positives, if you will, or assay specific type signals in the R3 through R12 channels. To validate the estrogen receptor bioactivity model, we needed to identify reference chemicals from other sources based on other data. So it wasn't an autovalidation, and we identified 40 in vitro estrogen receptor reference chemicals, independently confirmed using other assay sources and information sources. And these are, in large part, the same as the chemicals identified by OECD, the Organization of Economic Cooperation and Development, in 2012. And then we also identified 43 in vivo, or in life, estrogen receptor reference chemicals. Again, independently confirmed activity mostly from uterotrophic studies, so studies done in either rats or mice where there was an indication of uterotrophic activity that's estrogen receptor agonist dependent. And so we had 40 in vitro estrogen receptor reference chemicals independently confirmed and 43 in vivo estrogen receptor reference chemicals. A large pool of reference chemicals that will be used to validate the estrogen receptor high-throughput screening assays from toxic gas as well as the estrogen receptor bioactivity model integrating the results from those assays. The high-throughput screening assays very accurately detected the in vitro reference chemicals across this range of chemical structures. A broad range of potencies from strong chemicals, highly potent chemicals, to very weak chemicals and inactives. These reference chemicals have been peer reviewed, as I've mentioned already by OECD in 2012, as well as other international organizations, including ICCVAM. And a reference or URL here provides more information on some of that peer review for these in vitro reference chemicals. The estrogen receptor bioactivity model perform very well, identifying 26 of the true positives and 11 true negative compounds. It had a potential of two false negatives, but really, these were very weak chemicals, and it's arguable whether they're active or inactive. And so we're not too concerned about those two potential false negatives, very high degree of accuracy, 0.93% sensitivity, 0.93 in specificity 0.92. So the model performs very well against this external set of reference chemicals. Besides the in vitro reference chemicals, we wanted to identify a large pool of in vivo reference chemicals to see whether the in vitro assay and the pathway model could predict in vivo response, uterotrophic esatrogenic response. And we relied here very much on the uterotrophic assay to identify these chemicals, and then we work together with our colleagues at the National Toxicology program. And there, they identified or curated a large database of rodent uterotrophic results. And through a systematic review of the literature, to identify a range of chemicals and the range of chemical activities measured in this guideline-like uterotrophic studies and identify the in vivo reference chemicals. They had active chemicals verified in two or more independent studies, and inactive chemicals verified in two or more independent studies with no positive results in any study. So a good robust set of positive or active, and negative or inactive chemicals identified from this uterotrophic database. 43 in vivo estrogen receptor reference chemicals were identified in the publication in IRNTP colleagues. And the estrogen receptor model did a very good job accurately identifying these in vivo reference chemicals identifying 29 true positives, 8 true negatives, potentially 5 false positives. But again, these were mostly well-understood and very low activity chemicals, indoor chemicals that were difficult to test for variety of physical chemical properties, either in vitro in the high-throughput screening assays. So close to 90% accuracy and a great deal of sensitivity and reasonable specificity in the performance of the estrogen receptor model. The most important is the sensitivity of the estrogen receptor model. It was as sensitive or more sensitive than the uterotrophic assay, and what we found is that the uterotrophic assay itself was self-contradictory. In many cases where there were more than one uterotrophic assay run on a chemical, you would see a mixture of positive or active and inactive results. So combining the 40 in vitro and the 43 in vivo reference chemicals, there's a total of 65 unique chemicals that we could make comparisons of the high-throughput screening data and the estrogen receptor pathway model against a variety of orthogonal in vitro or in vivo data to validate the performance of these alternatives. And identify a range of different types of chemicals, chemical structure in a range of different activity from very strong actives to very weak actives and a large number of inactive compounds. The ToxCast Model did very well against endocrine receptor screening program Tier 1 type assays, particularly Binding, the Trans activation and the uterotrophic assay that performed as well or better that this existing methods from Tier 1. As I've mentioned, we evaluated the model's performance and validated it against 65 unique reference chemicals. By comparison, the Tier 1 Binding was validated using only 23 chemicals, and a large portion of those were not consistent with the expected outcomes in the validation effort without binding assays. The Transactivation assay was validated using only 12 chemicals, and the uterotrophic only 7 chemicals. The estrogen receptor model, besides the reference chemicals, was in 100% agreement with the first 52 chemical results generated for EDSP List 1 chemicals. So there were 52 chemicals beyond the reference chemicals that had EDSP Tier 1 data generated within the program, and we had 100% agreement. Amazingly good agreement between the ToxCast data for 49, almost 52 Pathway model against Binding and the Transactivation and uterotrophic results. I've mentioned that the May well be more sensitive than the Tier 1 assays. It's maybe partly due to the redundancy of the assays, 18 different high-throughput screening assays and the technologies. I also mentioned that the results from the uterotrophic studies, in particular, indicated fair amount of conflicts within those data. In other words, in this plot, we see in chemicals that have more than one uterotrophic study, there were often a mixture of active and inactive results from one study to the next. So the comparison, if you will, the gold standard of a source that we're comparing these alternative results to is not really so gold. There's difficulties in interpreting the uterotrophic result, and you'll see that as we continue on in the program, I think. Making comparisons to other types of in vitro, and in particularly, in vivo or animal studies where, when they're run more than one time on the same chemical, they sometimes give conflicting results. There's a difficulty in the reproducibility of these types of assays. The results for the estrogen receptor screening model, the high-throughput screening results indicated about 100 chemicals with estrogen receptor agonist activity out of total of over 1,800 screened. And those results are published in the Browne et al publication in 2015, and Environmental Science & Technology, as well as in the Federal Registry Notice where those results were published as well. The estrogen receptor agonist and antagonist activity identifying, as I've mentioned, about 100 of the 1,855 screened that had appreciable estrogen receptor bioactivity. The 18 high-throughput screening assays from ToxCast detected receptor interactions at various points along the signaling pathway. I'd mentioned that it's a mathematical predictive model. It integrates the area under the curve for the 18 different assays to give a single bioactivity value, very amenable to ranking the chemicals one against the other. It uses a variety of assay technologies. It does a very good job in distinguishing true activity from cytotoxicity or other false activity signals and gives a range of values for both estrogen receptors agonism as well as antagonism activity. We then went on to validate the estrogen receptor model using comparisons to in vitro assays as well as in vivo assays, particularly the uterotrophic assay, and then publish those results in a series of peer reviewed journal publications. Mentioned earlier in Toxicological Sciences, Environmental Health Perspectives, and Environmental Sciences &Technology. This ends this section introducing and describing the estrogen receptor pathway model. We'll move on now to talking about adverse outcome pathways.
