Extend the content, functionality and usability of the online FSK-Web model repository

The transfer and re-use of existing food safety knowledge (ranging from experimental data and published mathematical models to risk assessment software code) is still a major bottleneck in the area of food safety risk assessment. On the basis of the lessons learned from the development of a prototypic QMRA food and feed safety model repository “FSK-Web”, during Specific Agreement (SA) 3, this Specific Agreement will work towards the development and operation of a production model repository that will include risk assessment models that can be created, uploaded, downloaded and executed online, based on user input on model parameters. The repository should benefit risk assessors in EU Member States, and therefore needs to be fully functional, optimized and error free, as well as populated with a variety of useful curated models. Regardless of the final platform that will host the repository, the link to EFSA’s Knowledge Junction (KJ), as developed during SA 3, needs to be retained and be fully functional. Detailed manuals and online resources (e.g. video tutorials) need to be developed to support the use of the repository and its linking to the KJ. Functional details, including annotation guidelines, terms of use of the repository, curation of the content, software and license issues, authorship and notification of the authors, versioning, disclaimers, liabilities etc. need to be addressed.

In December 2016, EFSAshort forEuropean Food Safety Authority and BfRshort forGerman Federal Institute for Risk Assessment stepped into a Framework Partnership Agreement (FPA) to work jointly on the development, improvement and dissemination of global food safety tools. As outlined in the Specific Agreement #7 of this FPA one specific objective was to work towards the development of a fully operational online model repository that includes features to create, upload, download and execute models with user-defined inputs for the predefined model parameters. The user interface of the so-called FSK-Web model repository had to be designed such, that it is user-friendly and usable on various end-user devices and web browser software. The developed “FSK-Web 1.0” platform now even contains additional services, e.g. to link models with EFSA’s Knowledge Junction (KJ) or to join up to four models in a graphical user-interface (GUI) online. Detailed manuals, video tutorials and an online e-learning course were developed to support the use of the FSK-Web repository. The project created three key outcomes: (1) An operational curated web-based food safety model repository, and a significant number of executable QMRA models in FSKX format relevant for risk assessment; (2) Improved usability and annotation of the various services offered by the FSK-Web model repository including manuals and online resources explaining the functionality and use of all aspects of the model repository; (3) Provision of the possibility to link different FSKX models to create new combined models, including the option to combine models programmed in different scripting languages (also execution of combined R / Python models supported). From a technical perspective, the FSK-Web 1.0 platform is based on three elements: (1) A public JavaScript-based website that provides an open FSK-Web portal with a model search, sort, filter, details view and download option. It also has a menu that provide links to the FSK-Web protected area services and relevant related resources. (2) A number of web services hosted on a KNIME Server instance where registered users can perform additional operations, like execution, creation, joining and editing of models. These web services were all developed with the open source KNIME Analytics Platform (https://www.knime.com/knime-analytics-platform) and the open source FSK-Lab KNIME extension (https://foodrisklabs.bfr.bund.de/fsk-lab/). (3) A number of websites hosted on https://foodrisklabs.bfr.bund.de/safetytools-efsa-sa2/ providing the overview on available FSK-Web functionalities and links to corresponding manuals and video tutorials. With respect to the repository content, significant efforts were made to establish a balanced curation policy that take into account the various constraints from the underlying technical infrastructure, IT security aspects, annotation and model quality aspects (e.g. MIRARAM compliance) and legal aspects (e.g. copyright and licenses). With the drafted FSK-Web curation policy, it will be possible to ensure that the FSK-Web platform and its content fulfill highest quality and service standards in the future. During the project it was further proven, that the generation of a high number of FSKX models is challenging, as even the conversion of already published models into the FSKX format requires significant efforts for the model creator. Specifically, for cases where a published model is only available in a non-supported software, e.g. @Risk, the FSKX model creation requires to re-implement the original model into one of the supported scripting languages (currently R or Python). Also for models, that were originally created in one of these supported languages, the FSKX file creation process requires usually some adjustments inside the original model code as well as dedicated efforts to provide high quality model metadata (e.g. including on all model parameters) and the creation of a meaningful simulation scenarios and a visualization script. The curation policy also recommends seeking support from the original model authors for the quality control of the newly generated FSKX files, which might lead to a lengthy conversion process. Another unique feature of FSK-Web (and the underlying KNIME extension FSK-Lab) is the possibility to join FSKX models graphically. In this project a completely revised version of the so-called “FSK Joiner” KNIME node was developed that provides the central joining functionality in all online joiner workflows and for users of the FSK-Lab KNIME extension. As the FSKX format defines that all relevant model parameters are classified as Input, Constant or Output these model parameters can be graphically represented in the Joiner GUI. Different parameter types are also labeled by different colors and at specific positions of the model icon, e.g. an Output parameter is always on the right side. The GUI also performs automatically several quality control steps, e.g. it prevents the assignment of values to Constant parameters or the creation of circular connections between models. The graphical joining of models should in any case be performed after careful thinking of the users so that a scientifically and technically correct joining is assured. User benefit here specifically from the fact, that the complete model annotation of all to-be-joined models is directly visible in the FSK Joiner GUI. The FSK Joiner GUI even provides users the option to enter small “conversion code” snippets on the connections to modify the value of a selected parameter before it is assigned to an Input of a downstream model. This allows making necessary adjustments when for example data types or parameter units differ between otherwise matching parameters. In summary, FSK-Web 1.0 can be described as a flexible model integration and exchange platform that can easily be adapted, extended and modified as it is built on the modular KNIME framework.