A novel neural network technique for analysis and classification of EM single-particle images

We propose a novel self-organizing neural network for the unsupervised clas sification of electron microscopy (EM) images of biological macromolecules. The radical novelty of the algorithm lies in its rigorous mathematical for mulation that, starting from a large set of possibly very noisy input data, finds a set of "representative" data items, organized onto an ordered outp ut map, such that the probability density of this set of representative ite ms resembles at its possible best the probability density of the input data . In a way, it summarizes large amounts of information into a concise descr iption that rigorously keeps the basic pattern of the input data distributi on. In this application to the field of three-dimensional EM of single part icles, two different data sets have been used; one comprised 2458 rotationa l power spectra of individual negative stain im-ages of the G40P helicase o f Bacillus subtilis bacteriophage SPP1, and the other contained 2822 cryoel ectron images of SV40 large T-antigen. Our experimental results prove that this technique is indeed very successful, providing the user with the capab ility of exploring complex patterns in a succinct, informative, and objecti ve manner. The above facts, together with the consideration that the integr ation of this new algorithm with commonly used software packages is immedia te, prompt us to propose it as a valuable new tool in the analysis of large collections of noisy data. (C) 2001 Academic Press.