This report was provided to the Society of Toxicology following my selection as one of the recipients of the Supplemental Training for Education Program (STEP) award in 2020.
While my PhD dissertation research taught me how to utilize toxicology and exposure information and conduct risk assessments, I did not learn much about computational techniques that are becoming increasingly important in a Tox21 world. In particular, I was interested in computational toxicology models like Quantitative Structure-Activity Relationship (QSAR) and Read-Across that can be used to support risk-based prioritization of chemical inventories and address information gaps in chemical-specific risk assessments. Fortunately, there is a computational toxicology service, TOXnavigation, that provides training in these techniques. With Society of Toxicology’s generous funding through the Supplemental Training for Education Program (STEP), I could attend their live-stream sessions and learn about this exciting field!
ToxNavigation offers a “Use and applications of QSAR and read-across” course which is split into five sessions over five days (12.5 hours total). The course (offered several times in a year) is restricted to five participants to allow for frequent interactions, and is taught by Elena Fioravanzo, a computational chemist/toxicologist with over 20 years of experience. Elena did a great job walking us through different concepts, models, and applications, while skillfully answering the many questions I had! Apart from being provided a theoretical understanding, we were also shown how to use several free and commercial software programs and were provided homework questions to allow us to gain a better understanding of the material. At the end of each session, we were provided the presentation slides and video recordings and had the option to attend virtual “office hours” to further discuss topics of interest.
QSAR models and ReadAcross techniques can both predict the biological activity of an untested chemical by using available data from other chemicals. QSAR makes predictions through structural similarities with the help of an algorithm and training set. ReadAcross is more subjective: the toxicity endpoint (i.e., biological activity) predictions made by a user have to be justified. Both are often used by regulators (for example, scientists in European Chemicals Agency and the U.S. Environmental Protection Agency) to assess and manage chemical risks, by manufacturing companies when developing and registering chemicals, and by pharmaceutical companies during the drug discovery phase.
So, what did the course cover? We started off by looking at chemicals, specifically their properties, structures, and identifiers, and were introduced to chemical databases like ChemTunes, ChemSpider, CompTox, and COSMOS. We learned how to calculate chemical similarity and molecular descriptors, and were instructed on structural alerts (i.e., functional groups that are associated with certain adverse outcomes) and on how to identify them. We were exposed to several QSAR models like the OECD Toolbox, VEGA, and Danish (Q)SAR Models, and learned to generate QSAR and ReadAcross predictions and assess their reliability. Finally, we were shown ways to deal with uncertainties and report findings. You can get more information about the course here: https://www.toxnavigation.com/on-line
One of my goals is to employ computational techniques to my future research and risk assessments, either to lessen the amount of testing needed or to reduce uncertainties in existing data. SOT’s monetary support will certainly go a long way in helping achieve my career goals. I wish the same for other graduate students: please check out this website (https://www.toxicology.org/education/st/step.asp) and apply. Good luck!