This project has two main objectives:
i) to transform the R pipeline developed by Van den Berg et al. (2019) into a user-friendly tool that can be applied to multiple aquatic taxonomic and chemical groups and can simultaneously be used to evaluate important modelling aspects for the application of trait- and lineage-based models, and
ii) to deliver a set of case studies and model validation studies that demonstrate how this tool can directly be applied for aquatic risk assessment purposes and to proof the validity of the constructed models.

To ensure that both objectives are addressed with equal importance the project is split into two separate phases. Phase 1 is concerned with objective i improving and evaluating the tool on both technical and scientific aspects whilst phase 2 is concerned with objective ii constructing case- and validation-studies that demonstrate the application and validity of the tool and the models for aquatic risk assessment purposes. The phases are subdivided into four working packages (WPs) which will be described in more detail in the section on ‘Geplande activiteiten’.

Since it is impossible to experimentally determine the sensitivity of all the species present in any ecosystem to all chemicals to which they can possibly be exposed we rely on cross-species extrapolation of chemical sensitivity. Such extrapolation requires on the one hand a mechanistic understanding of how a chemical can cause damage and on the other hand a mechanistic understanding of which factors cause that one species responds differently to a chemical than another species. Van den Berg et al. (2019) developed an R pipeline that provides predictive aquatic macro-invertebrate trait models for a diverse set of pre-defined Modes of Action (MOAs). This pipeline was later extended to also include relatedness-based predictors since other studies found that both traits and relatedness explained a unique part of sensitivity. Advantages of this pipeline include reproducibility (all analyses are 100% reproducible permitting clear communication of modelling decisions and outcomes) flexibility (the pipeline can be adjusted to different organism groups) and speed (models can be compiled in a few hours). However so far this R pipeline has been constructed for research purposes only and still requires fluency in R to make changes to its functionality and applicability. For instance adjusting the taxonomic focus of the models would require making changes in multiple parts of the R pipeline. Therefore further development of this pipeline by transforming it into a flexible tool would make it directly usable for incorporation into risk assessment and can additionally allow the assessment of important aspects influencing sensitivity models (e.g. exposure duration effect endpoints). Once the tool has been developed further and evaluated on important sensitivity modelling aspects the next step is to demonstrate how the tool can be applied in risk assessment.

1.1 Evaluation of model selection algorithms.
1.2 Transformation of R pipeline into a more user-friendly and flexible tool.
1.3 Evaluation of important sensitivity modelling aspects.
1.4 Tutorial on the new R tool.2.1 Collection of macroinvertebrate traits.

2.2 Improved macroinvertebrate models.
2.3 Selection of sensitivity-related fish traits.
2.4 Develop the first set of fish models.

3.1 First compilation of case-studies.
3.2 Workshop on case-studies.
3.3 Final compilation of case-studies.

4.1 Model validation with single species tests with invertebrates.
4.2 Data collection and model construction for non-lethal endpoints for invertebrates.

The main innovation of this project lies in the use and exploration of all toxicity data collected over time and the possibility to flexibly select evaluate and incorporate various important aspects of these data into statistical models. This combination of flexibility and big data will assist in obtaining new mechanistic information to understand the sensitivity process better which can strongly stimulate fundamental research into the ecological and chemical process of sensitivity as well as contribute to the transition to a zero animal-testing society.First we explain the activities performed under this project along with their type (industrial fundamental or experimental) and the role of the involved parties within these activities. Next we provide two tables containing more detailed information on the budget spend on each activity and a calculation of the percentage PPS financing.

1.1 User-friendly and flexible R tool.
WENR, Unilever, other companies, risk assessors, regulators, policy makers will benefit from this tool.
The maintenance of the tool will lie with WENR (i.e. quarterly check on functionality and dependencies). Further development of the tool (i.e. incorporation of new functions) can be done by WENR if further funding will become available, but can also be picked up by any other party that is interested, since the tool will be published open and for free. Unilever will incorporate the tool in their risk assessment framework, and will help stimulate other companies to do the same.
1.2 Tutorial on the new R tool.
Anyone who will use the tool (see previous) will benefit from it. WENR, will check quarterly whether the tutorial and the tool are both still accessible and compatible with dependencies.
1.3 Journal publication on the evaluation of important sensitivity modelling aspects.
Scientists, risk assessors, regulators, modelers will benefit. Collaboration between WENR and Unilever. Usage is secured by open access publication.
2 Journal publication(s) on the improved tool, using the macroinvertebrate and fish studies as examples.
Scientists, risk assessors, regulators and modelers will benefit. Collaboration between WENR and Unilever. Usage is secured by open access publication.
3 Workshop publication on case-studies.
Scientists, risk assessors, regulators and modelers will benefit. Collaboration between WENR and Unilever. Usage is secured by open access publication.
4.1 Journal publication on the validation exercise.
Scientists, risk assessors, regulators and modelers will benefit. Main responsibility lies with WENR. Usage is secured by open access publication.
4.2 Journal publication on the development of predictive models for non-lethal endpoints for invertebrates.
Scientists, risk assessors, regulators and modelers will benefit. Main responsibility lies with WENR. Usage is secured by open access publication.

1.1 WENR Open-source software, likely as R package or Shiny tool. Source-code will be made publicly available on github.
1.2 WENR As an Rmarkdown vignette, either inside the R package, and/or on a website.
1.3 WENR Open-access publication(s) in peer-reviewed journal.
2 WENR Open-access publication(s) in peer-reviewed journal.
3 WENR Open-access publication(s) in peer-reviewed journal.
4.1 WENR Open-access publication(s) in peer-reviewed journal.
4.2 WENR Open-access publication(s) in peer-reviewed journal.

We will publish basic project information on the website of the TKI Deltatechnology.
We will also construct our own website, on which we will post i) blogs on computational problems/solutions we come across during the project, ii) links to publications/other projects closely-related to the project, and iii) any news directly related to the project (e.g. publications produced by project partners, information on the planned workshop). We will publish a link to our website on the website of the TKI Deltatechnology.

We will actively look for a direct collaboration with the Amsterdam eScience centre (https://www.esciencecenter.nl/collaborate/) for in-kind assistance with the computational structure of the project, and big data issues we expect to run into. We also expect to form some collaborations with other (trait-based) researchers, in order to obtain access to and assistance with state-of-the-art trait databases.
Also, we are open to any other partners, for instance policy makers, regulators, and other commercial companies. However, we will not actively look for them.