This workflow executes the testing pipeline for the MethylNet package. MethylNet is a modular deep learning workflow for embedding and prediction on DNA methylation data. In this example, we have subsetted the samples and CpGs of our age estimation set as introduced in https://www.biorxiv.org/content/10.1101/692665v1 and https://www.ncbi.nlm.nih.gov/pubmed/31368477, and then run a small set of randomized hyperparameter scans of the neural network topologies and other associated hyperparameters to predict age. First embedding model is tuned to yield the lowest validation loss, and then prediction layers are added for additional fine-tuning of the neural network. The real time training and validation losses of the models are recorded and output from the analysis are training curves for embedding and prediction tasks, interactive visualizations of the unsupervised and fine-tuned embeddings, plots depicting concordance between the predicted ages and actual ages. The pipeline depicted here runs using CUDA GPU acceleration. You can find the code on GitHub at https://github.com/Christensen-Lab-Dartmouth/MethylNet, on pypi at methylnet, and docker under the tag joshualevy44/methylnet.