Analysis of structural variability within two-dimensional biological crystals by a combination of patch averaging techniques and self organizing maps

We study in this work the use of self organizing maps to analyze the structural variability that can be found along two-dimensional crystals of biological macromolecules. Small areas of the crystals, termed “patches” by previous researchers, are used to obtain local average images that are then used as the input of a Self Organizing Map. This procedure allows for a fast and accurate image classification. Multivariate Statistical Analysis is then used on the resulting code vectors producing a very condensed data representation. This methodology is applied to previously studied crystals of bacteriophage φ29 p10 connector, finding a crystalline heterogeneity probably associated to multilayers in some areas of the crystal.