Nowadays it is becoming more and more common that our capacity to produce raw data is greater than our capacity to analyze them; this happens, for example, in genomics. Structural determination, and in particular electron microscopy, is no exception. Many images in many different settings can be acquired about a given specimen, and the questions then emerge: "What are all these pictures telling us about the specimen?" and "Can we find relationships among the images that convey us new information?".

I will concentrate on the Life Sciences, analyzing in particular the case of three-dimensional electron microscopy (3D EM) of non-crystalline specimen, where the analysis of tens of thousands of very noisy images is required in order to get an estimation about the three-dimensional structure of the specimen. Robust pattern recognition approaches are needed in the 3D EM field. I will review some of them, presenting a number of novel contributions that we have developed within the Biocomputing Unit of the National Center for Biotechnology in Madrid.