The MouseAge project was launched last year after successful fundraising on Lifespan.io. It aimed to develop the first photographic biomarker of aging in mice to help researchers assess potential anti-aging therapies, reduce animal testing, and speed up the pace of aging research.
The major advantage of this approach as opposed to other biomarkers of aging, such as the epigenetic clock, is that it is less invasive and easier to perform, thereby simplifying data collection for scientists.
The MouseAge team has been hard at work, and its project was recently featured in The Scientist, in which a number of researchers discuss how useful this free app could be in the lab. If you are a researcher and would like to use the free MouseAge application, you can visit the Apple Store and download it here.
Today, the project team has given us an update on the project and its progress, which we would like to share with you.
Milestone 5 – Gain collaborations
The following R&D collaborations have been agreed:
- The Buck Institute for Research on Aging, CA, USA (Prof. Eric Verdin).
- Syktyvkar State University (Prof. Alexey Moskalev and Prof. Mikhail Shaposhnikov).
- Charles River lab mouse supplier.
- Moscow State University.
Milestone 6 – Collect researchers’ feedback
The MouseAge team collected the users’ feedback to improve the app’s user experience and performance. Most researchers found the app comprehensive and user-friendly. However, they indicated what could be improved:
- Implement an algorithm for automatic background detection and deletion, as it is sometimes tricky to find a place with a uniform background to make the picture quality suitable for using the system.
Milestone 7 – Data quality assessment
- We implemented 4 algorithms for quality assessment of automatic images: movement blur, optics blur, sharpness, and exposure.
- After we collected the first batch of images, we performed the data quality analysis. We processed images with quality assessment algorithms and also used manual data quality assessment.
- Despite the quality of the images themselves being quite good, the standardization was not enough. We are now considering methods to improve standardization of images to make image assessment better.
- In total, we collected approximately 700 images of mice so far.
- We need to collect more data, as the ages of the mice in the images collected so far is not evenly distributed among age groups. More images and a more even cross-section of ages will improve the quality of the data.
Milestone 8 – First NN model development
- We developed the first neural network model for age prediction. The data was not enough to solve the regression task to predict the exact age. So, we decided to develop a classifier of ages. The model was aimed at predicting three age ranges – young, mature, old. The accuracy of the developed model is 0.67 (random choice is 0.33).
Milestone 9 – Set up a lifetime study of mice
- The MouseAge team is working with the Moscow State University mouse facility to collect the data of 50 mice from their birth to death; this will be a longitudinal study. This should be very valuable and help us to further refine the application and its accuracy.
Milestone 10 – Improvement of data storage and mobile app
- The MouseAge team is working on the improvement of backend architecture of the server which receives the data from the mobile app. We plan to write automatic tests and a more advanced error logging protocol.