Electron microscopy produces large amounts of extremely noisy image data - many people claim a signal to noise ratio of about 1:1. We work on computational methods to extract useful information from these signals using methods from machine learning and bioinformatics.