Leveraging data science to promote sustainable practices in severity assessment

• Applicant: Dr. Steven Talbot
• Subject Area: Sensory and Behavioural Biology ; Bioinformatics and Theoretical Biology ; Medical Informatics and Medical Bioinformatics
• Term: since 2024
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 321137804

In our project, we present an innovative data science project aimed at improving animal welfare in laboratory animal research. By implementing structured data flow and management procedures, we aim to measure the field quantitatively and establish robust severity patterns ('digital fingerprints') as features for impairment patterns applicable in real scenarios, such as for predictive purposes in laboratory animal research. To achieve this, we’ll use our Relative Severity Assessment (RELSA) score, which enables quantitative severity assessments at the individual or animal model level and will be applied to previously collected data from the research group. These efforts will evolve into a 'severity map/landscape', coding for the multidimensional topology of impairment features using RELSA and its weightings. This approach will enable risk assessments, severity mappings, and predictions for future and ongoing evaluations. Our plans emphasize scientific rigor and consider the diverse conditions in laboratories. We will use two primary data sources for the planned analyses: experimental data from the research group’s previous funding periods and telemetry data from a multicenter study. In addition, we can address specific questions by employing advanced computer-assisted analysis tools, such as our 'Severity Toolbox', e.g. for predicting humane endpoints. Additionally, using the data from the multicenter study, we plan to evaluate the robustness of severity parameters across different laboratories and validate the newly developed Smart Home Cage System. These approaches will consider the heterogeneity of experimental conditions to obtain realistic and robust estimators. Data such as heart rate, heart rate variability, temperature, and activity from transmitters, as well as body weight of mice and rats, will be used. We plan to test the non-inferiority of the cage system compared to invasive telemetry. Multivariate methods, mixed models, machine learning, and RELSA will be employed in these approaches. Furthermore, the project involves sensitivity and meta-analyses, for example, to evaluate severity parameters and their combinations. By exploring and utilizing the abovementioned patterns and their closely related topologies, we aim to achieve a more evidence-based and objective evaluation of severity assessment in laboratory animal research. Furthermore, this approach can provide researchers with more information about potential impairments to the animals and acute developments and, therefore, contributes primarily to improving animal welfare.

• DFG Programme: Research Units
• Subproject of: FOR 2591: Severity assessment in animal-based research