This editorial provides a brief account describing how the QSAR2021 came to be before highlighting the articles submitted as part of the proceedings of the meeting.
The 19th International Workshop on (Q)SARs in Environmental and Health Sciences (QSAR2021) took place from the 7–9th June 2021. This 19th Workshop, in a long standing series, was very different for a number of reasons. The motivation for the hosts (US Environmental Protection Agency (US EPA), International Council of Chemical Associations (ICCA) Long-Range Research Initiative (LRI) and American Society for Cellular and Computational Toxicology (ASCCT)) was to broaden the scope of the meeting beyond (Q)SARs such that it encompassed other so-called New Approach Methods (NAMs) including high throughput and high content (HT/HC) screening methods as well as the range of computational methods used to develop predictive models. The workshop aimed to bring together scientists from across the world along with developers and users of predictive models to foster discussion and debate around the practical implementation of these approaches.
The meeting was due to be held in person in June 2020 in Durham, North Carolina, US and as the abstract deadline approached at the end of January 2020, the number of abstracts exceeded expectations promising an exciting scientific programme. Then the COVID-19 pandemic took hold, and in the interests of the safety and health of potential participants, the only course of action was to postpone the workshop to the following year. After re-negotiating with sponsors, vendors alike, planning of QSAR2021 recommenced once more. However as the pandemic continued to wear on, it became increasingly evident that an in person meeting was still not going to be a viable prospect even in June 2021. As such, the hosts regrouped in late summer of 2020 to imagine how the QSAR2021 could be converted into a virtual meeting. There were many considerations at play: What software platforms would be needed to manage the different platform, poster sessions, how to enable real-time exchanges, how to manage the timezones, how to provide an experience that was as ‘in person’ as it could be in a virtual world.
Once a plan emerged, announcements were made on the meeting website and communicated to those who had submitted abstracts for 2020. Three platform tools were to be used– a collaboration tool (Discord (discord.com)) to facilitate real time chat, share announcements, Zoom (zoom.us) for the platform sessions and Remo (www.remo.co) for the poster sessions. The meeting was scheduled across 3 days and the platform sessions were run in parallel tracks to allow more presentations to be made on a compressed schedule. Platform presentations were recorded ahead of time to ensure time keeping on the day and platform recording links were provided to all delegates to allow for playback for those wishing to attend a different platform track. Poster sessions in Remo facilitated better real time exchange since the platform was set up with virtual tables that allowed posters to be shared on a virtual whiteboard. This was an important factor for the hosts given feedback on how other meetings had not necessarily accomplished a satisfying poster experience. Registration cost was another consideration to encourage attendance from students and researchers worldwide. Regular registration was set at $150 and students at $50 (for perspective, the regular in person cost had been set at $500).
Abstract submission was re-opened in January 2021 and by the deadline, we exceeded the number of abstracts even from 2020. Registrations for the meeting itself exceeded 350 attendees. This higher than expected number was due in part the accessibility of a virtual meeting and the nominal cost of attendance. Although there was a large presence of registrants from the USA and Europe, notable were registrants from China, S Korea, Brazil, Algeria and Japan.
In keeping with the tradition of QSAR meetings past, there were several keynote presentations. Three scientists were invited to share their insights and expertise. Professor Mark Cronin, Liverpool John Moores University, UK led the keynote on the first day of the meeting providing a perspective on how the meeting and the QSAR field had evolved over the decades and what opportunities and impact QSAR approaches would still offer in the future. His presentation entitled “Did QSAR Save Toxicology? And What’s its next trick? was a thought provoking perspective of how QSARs have shaped and impacted chemical safety assessments. Dr Rusty Thomas, US EPA presented ‘Can toxicology by any NAM be as sweet?’ to highlight the challenges in chemical safety and how the US EPA is addressing these with its comprehensive New Approach workplan which integrates regulatory, scientific and communication strategies and objectives. The third keynote presentation was given by Dr Elizabeth Mannshardt, US EPA who had been invited to provide a Data Science perspective to highlight the skills and competencies that new and experienced researchers need to educate themselves with to progress in an increasingly data centric world. The data volume and complexity in using NAM data brings with it many informatic challenges to overcome.
Also in keeping with QSAR meetings of the past, there was an awards session to recognise young, seasoned and lifetime researchers in the field. Professor Ovanes Mekeynan, LMC, University ‘Prof As. Zlatarov’ deservedly won the Lifetime Achievement Award. His contributions to the field spanning over 40 years have been immense though perhaps he is most familiar to current researchers as the technical lead for developing the OECD QSAR Toolbox. The mid career award was won by Dr Grace Patlewicz, a research chemist at the US EPA and one of the hosts of the QSAR2021 meeting. The promising early career award was won by Dr Kamel Mansouri at the US National Institute of Environmental Health Sciences. Short biographies are still accessible from the meeting website https://www.ascctox.org/qsar2021/qsar-2021-awards.
There were 6 broad themes discussed at the meeting:
1) Development, Evaluation and Application of QSARs to fill data gaps
2) Cheminformatic approaches to big data and biological activity profiling
3) Development, Evaluation and Application of QSARs and TTCs
4) Application of Tools
5) Non targeted screening, characterising uncertainty and informatics
6) Emerging issues
The articles captured in this special capture draw from 5 of these 6 themes.
As part of theme (1), Akturk et al1 described the development of QSAR models for the prediction of comedogenic potential of cosmetic ingredients. Using a dataset of 121 cosmetic ingredients and calculated molecular descriptors, the feasibility of developing classification models to predict comedogenic potential were evaluated using a selection of different machine learning approaches. The best performing models relied on a random forest approach.
As part of theme (2), Hines et al2 described how a generalised PBPK model could be used to extrapolate plasma kinetic profiles of a case study substance, nicotine, between species using in vivo and in silico data after accounting for differences in exposure routes. In silico-informed model parameterisations provided similar results to rat in vivo-based parameterisations, highlighting the potential use of in silico approaches when data are limited.
Yap and Raymer3 described a locality-sensitive deep learner approach to toxicity prediction. This is a deep neural network with attention mechanism and an optional instance-based feature weighting component. The approach was applied to both a synthetic dataset as well as other bioactivity/toxicity datasets including Tox21, androgen receptor activity and acute oral toxicity that had been previously published in the literature. In most cases, test set performance using a combination of a locality-sensitive deep learner approach in conjunction with a feed-forward neural network showed marked performance over individual models.
Khan et al4 who contributed to theme (3) described the development of QSTR for toxicity prediction in fish and algae for cationic polymers in addition to interspecies models in daphnid, fish and algae.
Cross et al5 described progress of an adhoc workgroup comprising researchers from industry and academia that was established with the goals of addressing a number of scientific and regulatory issues including: the identification of N-Nitrosamine mutagenicity and carcinogenicity reaction mechanisms, collection of all relevant experimental data, development of structure-activity relationships (SARs) consistent with mechanisms, identification of N-Nitrosamine carcinogenicity potency categories from SARs, and more precise methods for calculating acceptable intake limits for N-Nitrosamines based upon mechanistic analogues.
Gong et al6 described mechanistically driven identification of novel structural alerts for mitochondrial toxicity as part of theme (3). Eleven alerts associated with different mechanisms of action were developed using a combination of maximal common substructures and Bayesian statistical analysis.
As an illustration of tools as part of theme 4, Abedini et al7 showcased the Integrated Chemical Environment (ICE) which provides a user-friendly means of accessing and using a range of different NAM data and tools to contextualise and explore chemical bioactivity and molecular property information.
As an example of a contribution to theme 6, Bossa et al8 described the progress of an ongoing case study in the EU H2020 Gov4Nano project regarding the reuse of in vitro Comet genotoxicity data, focusing on the issues and challenges encountered in applying the Findable Accessible Interoperable Reusable (FAIR) principles for good data management and stewardship.
The handful of articles here only provides a snapshot of some of the contributions to the meeting. That said, overall the meeting highlighted exciting developments in the development of new models and tools and how predictive tools are continually evolving to address the different data streams now encountered in chemical safety assessment.
To make QSAR2021 a reality was the result of a lot of hard work from the hosts, the extended organising committee and the delegates themselves. Special thanks go to Rick Becker (American Chemical Council), Kristie Sullivan (Physicians Committee for Responsible Medicine), Canden Byrd (ICF), Nora Aptula (Unilever), Grace Patlewicz (US EPA), Gina Hilton (PETA), Mark Cronin (LJMU) and Judith Madden (LJMU) who worked tirelessly in the organisation of the meeting over the extended timeframe. Overall, it was a success despite the pandemic and showcased what was possible. The next meeting will be held in person in Copenhagen, Denmark from June 5–9, 2023. Look forward to continuing the discussion around NAMs and what exciting science will be shared at the next meeting.