UCLA

Studies of canid epigenetics can shed light on the relationship of the DNA methylome and mammalian longevity. The strong link between human and domesticated dog biology motivates the use of canids as a model for studying aging. Previous studies have examined factors such as age, sex, and weight and their associations with targeted bisulfite sequencing data using multivariate machine learning models. Creating visualization tools of these models facilitates community use of these resources. In this thesis, we describe an R-based web application platform (Shiny) that executes regression and classification models designed using buccal swab data obtained from 217 samples from previous studies. We developed CanidPredict, which allows users to upload their individual samples using either a CGMap file or a calculated methylation matrix and view predictions regarding age, sex, weight, sterilization status, and behavioural status as well as the general model performance for each trait. In addition, the goal of CanidPredict was to also create an application design which is user-friendly, can be used both on a server or locally, and can be scaled up to include further raw file formats and additional trait associations. CanidPredict is available at https://singlecell.mcdb.ucla.edu/DogAge/.