Pain is a complex combination of physical, emotional, and personal experiences. The Common Fund Knockout Mouse Phenotyping Program (KOMP2) is trying to better understand experiences of pain. The researchers are part of an ambitious project to genetically silence – or “knockout” – and characterize all genes that code for proteins in the mouse genome. This effort to generate "knockout mice" for every protein-coding gene is the first step before systematically carrying out a range of tests to understand each gene’s biological function, or “phenotype.” As part of a new effort, assays that incorporate machine learning were carried out to uncover genes involved with the experience of pain. Learning more about these genes in mice can be used to better understand how similar genes may cause or contribute to the experience of pain in humans. A better understanding of the underlying biology of pain in humans may lead to new treatments or assessments for pain, and may even help reduce our reliance on opioids.

Currently, it is hard to understand pain in pre-clinical animal model systems, like mice, and available methods often rely on studying a quick response to pain, not behavior over a long period of time. These methods are often lacking in their similarity to the human experience of pain. Because animals, like mice, can’t be asked how they feel, observation of the behavior of individual mice is often required. This approach is time consuming, labor intensive, and requires many mice. To overcome these challenges, KOMP2 researchers developed a machine learning scoring system that combines accuracy, consistency, and ease of use to provide a standard assay to study pain in mice that is also feasible for large-scale genetic studies.

During the validation of the machine learning processes the researchers analyzed 111 short videos, with a wide range of timing relative to pain experience and followed long term behaviors. They capitalized on innovations in machine learning to allow for accurate classification of specific mouse behaviors. They found they were able to easily score over 80 hours of video and find differences between mice in both response time to pain experience and breadth of behaviors. A video-based automated system is readily adoptable by other labs because the experiments are already typically recorded on video. The automated scoring feature adds an important refinement in examining behaviors indicative of pain in mice. It also improves reliability and speed with which the assay can be performed and because videos can capture a lot of behavior in individual mice, the approach can potentially even minimize the number of mice needed for a study.

Reference:

Machine learning-based automated phenotyping of inflammatory nocifensive behavior in mice. Wotton, J. M., Peterson, E., Anderson, L., Murray, S. A., Braun, R. E., Chesler, E. J., White, J. K., & Kumar, V. Molecular pain vol. 16 (2020): 1744806920958596. doi:10.1177/1744806920958596.

Learn more IMPC landing page: https://www.mousephenotype.org/understand/data-collections/pain/ (link is external)