SenzaGen AB (SENZA) Earnings Call Transcript

Hello, and welcome, everyone, to today's webinar titled Advancing NAMs for Skin Sensitization Testing with a spotlight on point-of-departure, challenging substances and active ingredients. Today's agenda will include beginning with a presentation by me, Tim Lindberg, Key Account Manager and GARD developer at SenzaGen, where I briefly will go over the principles of the GARD skin assay followed by a few case studies to highlight how it can be used for testing of challenging test materials as well as looking into the GARD skin dose response for quantitative potency assessment. The second part will be a recorded version with invited guest speakers from Cargill, Unilever and Lundbeck. And then we will end with a live Q&A session, where you are free to ask any questions. Please do so in the chat box to your right. So to begin with an introduction to the testing for skin sensitization. Here are all of the key events that comprises the AOP for skin sensitization where we have the different in vitro assays assigned to the different key events, where you can locate the GARDskin assay and the Key Event 3 which was recently adopted as an OECD test guideline 2 years ago. And this provides a unique and mechanistically different method to monitor Key Event 3 in the AOP. And this slide shows the principles behind the GARDskin assay. We use a dendritic cell-like cell line and like other in vitro methods, we monitor the cellular responses towards sensitizers and non-sensitizers. However, unlike -- for instance, h-CLAT that looks at 1 and 2 or 2 or few biomarkers, the GARDskin assay monitors a larger set of gene markers in a biomarker signature, where we have the GARDskin comprising 196 genes. And if we look further into these genes that we here have mapped to the key events in the AOP for skin sensitization. It's evident that, of course, they target Key Event 3. We see common maturation marker like CD86 but also antigen recognition molecules like toll-like receptors. Additionally, we can also see that some of the genes in the GARD prediction signature also targets other key events other than Key Event 3. For instance, the NRF2 pathway in Key Event 2. And with this holistic approach, you get a broader sense of the information gathered from skin sensitizers. And this is something several of our clients use in nonregulatory settings to only use the GARDskin assay as a stand-alone assay. And here is a brief overview of the protocol for the GARDskin assay, where the binary hazard identification is used at only one single concentration where we stimulate the cells, extract RNA and this RNA is then used to construct the expression profiles of the 196 genes. The expression profiles are then used in a machine learning algorithm, in our case, a support vector machine and what's important to remember here is that all of the genes contribute to the final classification. So everything is put into one single value called a decision value and the interpretation of this is that if this decision value is positive, it's a skin sensitizer and if it's negative, it's a non-sensitizer. And as I mentioned previously, because skin assay was recently adopted as OECD test guideline. And for this, we performed a rigorous validation process, including a Ring trial that you can see the performance statistics from here. I will not go further into detail about this, but you are free to look into the published article about this validation process. What I want to highlight is and one of the most important things with regards to skin assay is that there exists quite a lot empirical data to support the claims of applicability for the assay in different domains. And here to the right, you can see some of the domains that we have explored together with partners and clients. And today, I would like to highlight and give a few case studies on complex mixtures and UVCPs. So the first case study is a collaboration between SenzaGen and Exxon Mobil where we were challenged with 16 UVCPs and formulation lubricant products. These are, of course, very hydrophobic and very complex materials, so they generally fall outside of the applicability domain, all provide more predictive value in other NAMs. The rationale behind why Exxon wanted to test the GARDskin was one, firstly, the availability of a large panel of alternative solvents that we can use and have validated for the assay. Secondly, it's also the notion I mentioned earlier that the gene signature aligned with multiple key events in the skin sensitization AOP, which could potentially reduce unnecessary testing in a nonregulatory setting. So looking at the results from this testing, we can see that 81% were accurately predicted and reflected the weight of evidence prediction from the references. Further, we can see that 2 out of the 3 incorrectly predicted test materials also had conflicting information or conflicting results in the references. So the conclusion drawn from this project was that GARDskin was effective and efficient for this very difficult to test materials such as UVCPs or formulations and it holds high concordance as compared to traditional weight of evidence approach. The second case study on the applicability domain for GARDskin is a collaboration between SenzaGen and Corteva Agriscience, where we investigated agrochemical formulations. And the rationale behind this testing was that the other in vitro methods, had poor predictive value for this set of chemical formulations as highlighted here in publications from 2022. Also, especially in the Key Event 3 domain, we can see that h-CLAT have very poor predictive value in terms of false positive generation, where 12 compounds of the 2 non-sensitizing were considered as false positives. So the testing we performed with Corteva consisted of a blinded testing of 42 liquid and solid formulations. And most of the testing was performed according to the established test guideline with the exception that we used an approximated nominal molecular weight of 400 grams per mol as, of course, many of these agrochemical formulations doesn't have a defined molecular weight. Looking at the results from this testing. We start with the liquid formulations and it's evident that the predictive performance is improved, where we have one false negative and five false positives. And before I go more into detail on these mispredictions. I also want to cover the overall performances when we also include the solids. And here, we can see that much of the performance is retained when also including these solid formulations. Going back to the false predictions that I've highlighted here in the table to your right, in yellow. We can see on the lower part, we have the true negative as in the reference data of LLNA. We can see that they, of course, don't have an EC3 value as they are negative but they have a maximum stimulation index that is close to the classification threshold of 3, meaning that they have some dose dependent manner in them. And it was also investigated and found that some of these compounds or these formulations were containing skin sensitizers. So all in all, we can see -- we theorize that some of these false predictions can be due to a sensitivity issue in the in vivo assays as compared to in the in vitro assays. So the conclusions drawn from these testing was that GARDskin demonstrate a high predictive performance for this kind of chemical domain that agrochemical formulations are that you also need to be very careful of how you choose and look into your reference data. Also, although I will not cover that in this presentation, we saw here that the predictive performance was increased when you use GARDskin as compared to h-CLAT. This is also true in a defined approach where predictive performance is increased if you use GARDskin as a drop in replacement of the h-CLAT method for Key Event 3 assays. The next part, I will cover the GARDskin dose response assay, which is a quantitative assessment assay. It uses the same protocol as the regular GARDskin. But instead of looking at just one single concentration, we use a titrated range of at least 6 concentrations to identify the lowest concentration that is required to induce a positive classification in the prediction algorithm. And we have, through different articles and a publication together with L'Oreal investigated how this cDV0 value correlates with human NESIL values where we through a simple linear regression model can make a prediction from the experimentally derived cDV0 value to predict a value in microgram -- in the unit of micrograms per square centimeters, which, of course, is the relevant unit when looking at skin sensitizing potency. And this can then be used further in any kind of available procedure for risk assessment. And here's just a brief example of how the assay is performed. First, we do the experimental testing to generate a dose curve. We plot the dose curve and identify the cDV0 concentration. We then use the cDV0 concentration in the linear regression model to identify a corresponding human NESIL or LLNA EC3 value, which, in this case, for the test item benzyl cinnamate becomes around 5,500 micrograms of -- micrograms per square centimeter of tissue. The first case study on this -- on the GARDskin dose response assay is a combination of several manuscripts that have been published where we, together with IFF, RIFM and L'Oreal, investigator reproducibility and performance of the GARDskin dose-response assay. So here, the predicted NESIL values from the GARDskin dose response assay is compared to a highly curated set of reference data on human NESIL. And we can see that the R-square value of this correlation is 69%, looking at the GARDskin compared to the human NESIL. What's also important to take away from this is that the same data when you compare LLNA versus the same human data, there R-square value becomes 60%. So conclusions from this is that the GARDskin dose response assay predicts human NESIL equally well or better than the LLNA data as compared to human data. The second part of this was the investigator reproducibility of the dose response assay. And here, we did repeated measurements of 27 compounds, where we had an average fold-change of 1.78 between the repeated measurements. So the conclusion from these projects were that the dose response provide an accurate and reproducible potency prediction and highly correlates with human NESIL values. The next case study on the dose response assay is a collaboration together with IFF, where they investigate if you can use a sensitizer in your product and if yes, what would be the maximum allowable concentration. So they performed next-generation risk assessment to identify the maximum allowable concentration for a sensitizer in different products. And for this, they used QRA2 framework. And here, we can see the iteratively different tiers that are used where you start with investigating in silico and read-across data. And if it's found that it's a sensitizer, you proceed with hazard identification in, in vitro methods in different labs. And if you still identify it as or you confirm it as a sensitizer you proceed with quantitative risk assessment in the GARDskin response assay. And here is the case for using this with isocyclocitral where in the north tier, we could see that it was predicted as a sensitizer further confirmed in Tier 1 in different NAMs. And then in the second tier, we identified a point of departure in the GARDskin dose response assay that was then used in the QRA2 framework. And together with different risk assessments, the table to the right was identified where you can see the different maximum levels that could be used in different consumer products for this compound. And with that, I conclude this part of the presentation, and we'll turn over to the prerecorded versions of our invited speakers. Please go ahead. Marie Tintin from Cargill, presenting the applicability of GARDskin for assessing skin sensitization potential of hydrophobic esters during product development. Marie has been a regulatory specialist in the cosmetic industry for 15 years with experience in both finished products and ingredients. She holds an MSc in cosmetic science as well as in toxicology and ecotoxicology. She actively supports the development of cosmetic ingredients by ensuring the safety for both human and environmental health. Please go ahead, Marie.

Hello. I am Marie Tintin. I am principal scientist in the Scientific and Regulatory Affairs department within R&D of Cargill since 7 years. I have a master's degree in cosmetic science and in toxicology. Today, I will introduce you to the work that we have done with SenzaGen to assess the skin sensitization potential of hydrophobic esters. Cargill is an agricultural company, which is delivering ingredients mainly to the food and feed market. Cargill Beauty, the division I am working for, is offering a broad and diverse portfolio of nature-derived ingredients to the personal care market. Our products are from botanical sources and are covered by different sustainability programs. You can check all the necessary information and download our portfolio right here. In this portfolio, you will be able to see texturizers, emulsifiers and emollients, among others, as Cargill is continuously developing new ingredients. When we develop new ingredients on top of efficacy and technical properties, we want them to be nature derived, COSMOS compliant, biodegradable and vegan suitable. Here, we were developing within 2 different projects, different substances. And once we have candidates that are good enough according to our requirements, we start to build the toxicological profile. Here, Substance A and Substance B are two esters representative of ranges, they come from vegetable oils from different botanical sources that have been esterified and/or transesterified to give complex compounds with variable composition from 12 to 80 carbons. Depending on the process that we apply, we can have liquid, pasty or waxy form. The products we have decided to test have similar chemical properties. They are both very hydrophobic, log of POW above 8. They have a very poor water solubility below 1 milligram per liter at 20 degrees. And besides the low water solubility, they are also very poorly soluble in the most conventional solvents as described in the OECD testing guidelines like DMSO, acetone and ethanol. So when we want to build the toxicological profile and apply different in vitro testing, therefore, we have a challenge for the solidization of our products. So here for the skin sensitization endpoint, we have decided to apply the 2 out of 3 testing strategy as described in OECD testing guidelines 497. We usually start with DPRA and h-CLAT, and here are the results for our 2 products. So for Substance A, DPRA came back negative while h-CLAT we have not been able to conduct the testing. The lab has tried to solubilize our ingredients in acetone, in ethanol, it didn't work. Our product was only soluble in THF, tetrahydrofuran. Unfortunately, the solvent cannot be used with h-CLAT testing. Therefore, we are pulling out of the applicability domain for this product. For Substance B, we have decided to test 2 different batches as we are still developing the product. For DPRA the results came back inconclusive. We have not been able to set a conclusion because for 1 batch, we saw a depletion of cysteine, but around 8 and no depletion in lysine. While for the other batch, we saw a depletion of cysteine at 10.3 and no depletion in lysine. So for us, we could conclude that 1 patch was positive, 1 patch was negative. For h-CLAT, we have -- we also decided that the results were inconclusive. We had, for one batch no increase in the marker expression, CD54 and CD86. While for the other batch, we had a slight increase of CD54 marker expression. We would conclude that h-CLAT would be not sensitizer, but we have still been intrigued by this gene expression that we could see. Plus the lab told us that they had also difficulty to solubilize our product. They managed to get a dispersion, but that was not that good quality. So we decided to disregard those results for inconsistency. Therefore, we have decided to conduct a third testing, the GARDskin assay. The lab has struggled as well to solubilize our products. For substance A, they managed to dissolve our product in acetone after heating and applying sonication technique. For Substance B, it has been more easily solubilized in ethanol. Here are the results for GARDskin assay. So here we can see that for Substance A that was solubilized in acetone, we have a decision value that's negative. And for substances B it has been dissolved in ethanol, we have also a decision value that is negative. Therefore, we can conclude that both of our products, non-sensitizer in GARDskin assay. But what about the risk assessment? What about the final conclusion? So for Substance A at this point, we can conclude quite easily that because we conducted 2 testing that came back negative. So h-CLAT and GARDskin, Substance A is non-sensitized. For Substance B, DPRA and h-CLAT are for us nonconclusive, only GARDskin assay is positive -- negative. So at this point, we decided to complete our assessment by applying further NAMs and new approach methodologies that include not only in vitro method but many other different methods. So here, this is an extract of the webinar of SenzaGen that illustrates very well which OECD method is linked to which key events. So here, we already applied DPRA and h-CLAT testing, and they are covering Key Event 1 and Key Event 3. We then applied GARDskin assay that is also covering Key Event 3. So at that point, if you want to pursue with further testing, we should use a testing that is covering Key Event 2. While the most used testing is KeratinoSens, but the lab we are working with at that time was not proposing KeratinoSens, but was proposing a similar testing that is not regulatory approved so that is used only for screening. And when we did the testing, the result was negative, which gave us enough confidence to pursue the assessment. And at that point, to have full confidence in our results, we should apply a testing that is covering Key Event 4. But we all know that we cannot use animal methods to assess the safety of a cosmetic ingredient. So the only in vivo method, we would be able to apply and that would cover Key Event 4 would be HRIPT testing. So that's what we did on 53 volunteers as we were also assessing the skin compatibility of some prototypes that we have also been developing. The results came back negative for this HRIPT testing. Because the testing we applied to cover Key Event 2 and also the testing we applied to cover Key Event 4 are not regulatory valid, we decided to further pursue this assessment by doing some literature and market search. We found good surrogates for both of our products, Substance A and Substance B that have more than 80% structure similarity with our products according to OECD toolbox. We validated it with QSAR. So here is the final assessment for our 2 substances. So by combining different NAMs and different techniques, starting with in vitro method for Substance A we have DPRA and GARDskin assay that came back negative. And this read-across with the surrogate came back also negative. We can, therefore, conclude that Substance A is a nonskin sensitizer. For Substance B, we had the GARDskin assay that was negative, the read-across that was negative and the HRIPT that was also negative. Therefore, we are confident enough to conclude that it's a nonskin sensitizer. Something else that we could do as well if we -- when we would, for example, to do the registration of our product would be to run a QSAR for Substance B to strengthen the toxicology profile for the skin sensitization end point. So that would demonstrate that by applying multiple nonanimal methods or new approach methodologies, we are able to build a strong risk assessment for our ingredients. Thank you for listening. I remain available for any questions.

Thank you for your presentation, Marie. Our next guest speaker will be Georgia Reynolds from Unilever with the presentation Skin Allergy Risk Assessment, the SARA model for GARDskin dose response as a possible input. As a science leader at Unilever, Georgia focuses on a applying exposure-led NAMs, including in silico and in vitro approaches to next-generation risk assessment for consumer product ingredients. Georgia's science leadership focuses on NAMs for skin allergy and application of integrated computational modeling approaches. Please go ahead, Georgia.

Hello. My name is Georgia Reynolds and I work for SEAC the Safety and Environmental Assurance Center in Unilever and Unilever is a fast-moving consumer goods company and because we sell a lot of consumer products, which may be dermally applied, skin sensitization risk assessment is an important part of our safety assessment. And today, I'm going to be talking a little bit about that model and how the GARDskin dose-response assay might be a possible input into that. So SARA as a model is a Bayesian statistical model. It's used within a wider next-generation risk assessment framework and it produces 2 main metrics. So one is the point of departure called the ED01, which is a 1% sensitizing dose in a human population for a chemical of interest. And the second metric is the risk metric, which is a probability that a consumer is exposed to a chemical is low risk. So the model was first published back in 2019, and then it underwent an extensive evaluation and the database was revised and expanded. And then we published on that SARA model in a set of 3 papers, which can be found by the QR code at the end of my presentation. Since 2021, we've been developing the next version of SARA. And this version of the model is described briefly in the table at the bottom of this slide. So the SARA model now has a database of over 400 chemicals and has a range of assay inputs such as historical data, including the LLNA mouse studies to human data, the HRIPT and HMT data. But primarily, the SARA model is used for new approach methodologies data or NAM data. And so a number of the OECD test guideline assays for skin sensitization are utilized such as the KeratinoSens, h-CLAT, U-SENS, the DPRA and kinetic DPRA. Most recently, we have added reactivity classifications into the model and they're used as priors. So we have classifications, which are determined by in silico and expert analysis and cover nonreactive autooxidation possible reactive and high potency chemical classes. As I mentioned, the model produces a risk metric, and this is determined using a number of risk benchmarks and these chemical exposures are defined as either high or low risk depending on the clinical data that is available for them. And as I mentioned previously, also provides an ED01 and the probability that chemical is sensitizing or nonsensitizing. And the final sort of addition to the SARA model in the recent updates has been we've moved to a version of the model called the production model, which allows us to run predictions in a matter of minutes instead of hours as previously. So now we're looking at whether the GARDskin dose response might be a viable input assay for this model. As we've heard, the GARDskin dose response assay is an in vitro test for quantitative skin sensitizing potency assessment and is built on the GARDskin assay and uses the same 196 transcripts. The assay provides an estimated threshold concentration or the CDV0 and for test substance to induce skin sensitizing effects. On the right-hand side of this graph, there's an example of how this is done for trans cinnamic aldehyde. And as you can see, there are 6 test concentrations and then a linear interpolation for defining the CDV0 value. So the SARA model assumes correlations between its assay inputs. And an example of this is the KeratinoSens assay, EC1.5 and ED01. Our hypothesis here is that the CDV0 value can be used in a very similar way and hopefully correlates the same way with ED01. So we began working with SenzaGen on a small number of chemicals, and we wanted to make most of these in the small panels. So we wanted to cover range of potencies, use some benchmarking chemicals and also look at a few chemicals, which have had interesting data in the past, either for the GARD assay itself or in other NAMs. And so just pull out a couple of examples there. benzyl alcohol, for instance, has been surprisingly potent in the past in the GARDskin dose response. And for squaric acid other NAMs have underestimated disclassification previously. So we've put a few of those types of chemicals in as well. We've also got a number of comparative CDV0 values, which are also modeled. These values are sourced from the paper on the slide published in 2021 and covers 29 chemicals with varying potencies. And the important thing here for the 29 chemicals is that even if a CDV0 value wasn't reported, we're still able to model that using the sunset data as the maximum concentration tested. So starting to look at the results. And on the left-hand side, you can see a graph on the y-axis of this graph is the CDV0 value and on the x-axis is the expected ED01. As you can see, for the chemicals that we've tested and for the published data, we see a very good correlation. And so in blue are the published values and in yellow are the Unilever generated data and the sunset values that I spoke about where you get the maximum dose tested, and you see upward-facing triangles and that's the sunset data. The dotted line is theoretical maximum in a HPPT study. So on the right-hand side of that line, that's where you would expect to see your nonsensitizers with CDV0 values above the maximum dose tested. And as you can see, the correlation is slightly stronger on the left-hand side of the graph, where you have more potent chemicals and it does slightly weaken as you get closer to that line. And you do get a few of those sunset values just slightly to the left of the line and where the maximum dose tested is above the CDV0.. So we do also have a number of outliers. And squaric acid, for example, is coming out as a CDV0 value above maximum dose tested. TDMS, again, an outlier are at bottom of the graph and benzyl alcohol you can see is a sunset value from the Unilever data, but previously when it was tested was more potent. So I'm going to talk about those a little bit more. So thiram is an organosulfur pesticide, and it has an extremely low CDV0 value relative to its in vivo potency estimates. So in an LLNA, it has a EC3 of about 5.2%. And in the HPPT study, a 15,000 micrograms per centimeter squared. I'd just like to point out here that this isn't a GARD-specific outlier, but a NAM outlier. So we do see similar results with other NAMs where you get disagreement between the in vitro data and the in vivo data. So just for example, you see 100% cysteine depletion in the DPRA and the kDPRA being very reactive. We don't really know why this is. So we're going to continue to kind of look into why the NAMs might be producing a different results of thiram. As I mentioned, squaric acid is also an outlier. We know that squaric acid is a strong sensitizer because it's used for the treatment of alopecia and triggers allergic contact dermatitis to redirect the inflammatory response as treatment. But in the GARD dose response, we see a CDV0 greater than the maximum dose tested. It has an LLNA EC3 of less than 2.5. But again, the human potency is not reflected by the other NAMs. So this, again, is not a GARD specific outlier. And we have things such as the DPRA being around 50% depletion, but no reactivity rate and the KeratinoSens assay is negative for this material. And finally, benzyl alcohol. So as I mentioned previously in the published literature, benzyl alcohol has been fairly potent in the GARD assay. So we were interested to know whether we could reproduce that result but when we repeated it, as you can see, it appears as an upward-facing triangle in yellow, and it was demonstrating that we expect benzyl alcohol to probably be a non-sensitizer, which aligns with the in vivo data. So as I've been talking about, we can use the maximum dose concentration tested, our sunset data if we don't define a CDV0 value. And in the same way, we can use the data from the GARDskin assay. So this is less informative than the data that you get from the dose response assay, but it is still helpful as an input into the model. So obviously, when you have a sensitizer, it could be any value less than the decision value that is determined through the GARDskin because there is only one test concentration, fairly high concentration. So there's quite a range of values that it could be below that, but it is still informative. And that's represented on the graph to the right-hand side with those green triangles. So in conclusion, potency estimates from the GARDskin dose response are baring a small number of outliers consistent with those obtained from the SARA model. And that suggests that the GARD CDV0 value could be a useful input. But at the moment, the model requires a fairly significant amount of data in order to be able to adequately model variability and we need more reproducibility data in order to be able to model it accurately. So this is something we're going to look at. And finally, and importantly, SARA is a weight of evidence model. And so it has a huge benefit over running any singular assay because it allows us to utilize the breadth of data and it will minimize the impact of any outliers as we have seen. So finally, I'd just like to say thank you very much to my team at Unilever and also thank you to the team at SenzaGen for all of their support throughout this study. And as I mentioned on my first slide, you can access via this QR code the Unilever website and for SEAC, which will have those papers for SARA and also this presentation, if you'd like to find it. Thank you.

Thank you for your presentation, Georgia. The next guest speaker and our last one is Camilla Taxvig from Lundbeck, who will present the application of the GARDskin dose response assay for assessing skin sensitizer properties of drug products and active substances. Camilla works as a toxicologist at Lundbeck, where she oversees the nonclinical safety evaluation of drug candidates. And for many years, she has been an active member of the OECD Validation Management Group for Non-animal Testing and Endocrine Disruptors, contributing to the validation and recommendation of OECD test guidelines focusing on NAMs. Please go ahead, Camilla.

So hello, everyone. My name is Camilla and I'm here today to tell you about Lundbeck's recent collaboration with SenzaGen where we did GARDskin dose response study with a drug product and 2 active substances. So we normally don't test for skin sensitization as a standard, but we do it on a case-by-case basis. And the reason for this latest study we did was that we have had some incidents of allergic reactions in one of our production areas. And of course, we take this very seriously. So we wanted to do some investigation to kind of see how we could resolve the issue. We had a specific drug product in mind that we thought might be the cause. So what we initially did was that we ran the drug product as well as the 2 active substances in the drug product in silico software. Specifically, we used the Derek Nexus software from Lhasa. And here, we ran a prediction for determining the skin sensitization potential of the 3 compounds. The results of the in silico prediction confirmed that all 3 compounds were likely to cause skin sensitization. So therefore, we wanted to confirm these prediction with some experimental data. And therefore, we contacted SenzaGen with the aim of running a GARDskin dose response study. And the reason why we chose the GARDskin dose response study and not just the GARDskin assay was then that in addition to actually getting confirmation of whether or not the compounds with skin sensitizers or not, we were also interested in knowing something about the potency of the different compounds and potential differences in potency between them. So here, I try to summarize the results that we got from the GARDskin dose response study. And as you can see, the study confirmed that both active substances and the drug product were classified or categorized as sensitizers but based on the result output that we got from the study, like, for example, the threshold effect concentrations and the predicted NOEL, we were able to classify the compounds and the 2 active substances were classified as strong sensitizers, whereas the drug product was classified as a weak sensitizer. The predicted human NOEL value represents the concentration above with a positive response for sensitization is estimated to be observed in humans. So if we look at the CLP classification criteria for skin sensitization based on positive responses in humans, then compound is classified as a category or given a Category 1A if the positive response is seen at or below 500-microgram per square centimeter. If the positive response is seen above these 500-microgram per square centimeter, then the compound is categorized as a 1B. So based on the human NOEL values that we got from GARDskin dose response study, we can conclude that both the active substances are Category 1A sensitizers, while the drug product is a Category 1B. And this is in line with what I showed before in the table where we based on the result output from the study, we're able to categorize the 2 active substances as strong sensitizers and the drug product as a weak sensitizer. So the results from the GARDskin dose response study showed that all the 3 compounds exhibited a clear dose-dependent increase in the decision value, so the endpoint measurements that are used in this assay. And based on the results that we got, we could conclude as already mentioned, that all 3 compounds could be classified as sensitizers but with different potency, which was one of the information end points that we were interested in actually getting information about. So not just confirming if there are sensitizers or not. So as mentioned, the active substances were more or less equally potent and were classified as strong sensitizers while the drug product were classified as a weak sensitizer. So in the context of comparing the effect of threshold concentrations, I just want to mention that the concentrations of the API or the actual pharmaceutical ingredient in the drug product is approximately 10%. So it could be an indication that the dilution of the active substances in things like [ CPMs ] and water in the final drug product is reflected in the effect concentrations of the drug product because as you might remember from the table, this effect concentration for the drug product was much higher compared to the effect concentrations of the active substances. So this could indicate that the sensitizing effect of the drug product is actually originating from the active ingredients for the 2 active substances. According to current EU regulation on classification and leveling of chemicals. So this CLP classification, which I already mentioned, then chemical must be classified as a sensitizer if there's evidence of sensitization either in humans or from results from animal tests. However, as the GARDskin assay is recognized as a test method under the REACH regulation and has also been accepted as an OECD test guideline, then it's actually a valuable method for getting reliable data for determining the sensitizing potential of compounds. So based on the results that we got from this study, we came out with some recommendation for the working environment. These recommendations are more or less in line with our standard recommendation. But it actually also shows now that we have gotten confirmation that compounds were able to be skin sensitizers that it's important to actually comply with these recommendations and make sure that you wear the proper protection, so proper working clothes and gloves and that you make sure to wash your hands when you have handled these compounds. And because in this situation, we're also talking about compounds that can cause allergy. They should not be handled in an open area, but in a closed system with a proper ventilation. So with that I just want to thank you for listening, and I look forward to taking any questions that you might have. So thank you.

Thank you to all our speakers for their presentations. They will shortly join us for the live Q&A where we also will have SenzaGen Chief Scientist, Henrik Johansson, join us. While we are setting up, I wanted to highlight our next event here at SenzaGen and also encourage you all to send in your questions to our panelists in the chatbox to your right. Additionally, feel free to send any requests on your in vitro toxicity testing needs to me, and I will be happy to assist you.

So let's see our panelists are joining. And there, all are welcome, Marie, Henrik, Camilla and Georgia. Great to see you all. Perfect. We'll dive right into the questions. I have one here directed at you, Henrik. In what situations should I currently consider GARD over other regulatory skin sensitization tests?

Well, of course, it depends on the specific needs and the reasons for testing. But from a R&D perspective, I would argue that GARDskin is a very well-suited method for screening of compounds, making prioritizations in R&D pipelines, either as GARDskin or GARDskin dose response use depending on the needs. In a regulatory context, of course, there are applicability domains to consider, but also, of course, gold standards regarding defined approaches and ITSs. And here, GARDskin is accepted as a weight of evidence data source. It can be used in ITSs if motivated appropriately, of course. Importantly as well, we are currently working on getting GARDskin data accepted as a data source also in the defined approaches of 1 and 7. So that is work ongoing. But the main reason I would argue in the regulatory context would be, for example, h-CLAT or any other Key Event 3 method is inappropriate of the specific applicability domain. And we have data supporting, for example, in hydrophobic compounds that GARD may be of an advantage in these cases. But there may be other reasons, of course. The best thing to do is, of course, to contact him and ask for meeting and we can discuss your specific purposes and your specific needs.

Thank you for that answer, Henrik. The next one is directed to Marie from Cargill. Since GARDskin assay gives info on genes that are involved in several key events, does it make sense to start with GARDskin before using other in vitro tests in a regulatory setting?

Thank you for that question. And actually, that's one of the first questions I asked the team when I met with Tim because my company has started to work with him even before the testing was validated by OECD, so a few years ago already. And actually, to this day, GARDskin assay cannot be used as a standalone testing to assess the skin sensitization endpoint. So we can start with this testing. Of course, it's very valuable information, but although it covers different key events, it's officially recognized as covering only Key Event 3. So what we can do today is to replace, for example, in our testing strategy, h-CLAT testing by GARDskin assay. But for regulatory purposes, we would still need to conduct either DPRA, either KeratinoSens as other testing to have an assessment that would be sensitizer or non-sensitizer.