Important numerical features for novel endoplasmic reticulum genes classification of protein localizations in micrographs

Abstract

Localization of proteins ties closely with the protein functions-the disorder of protein delivering can cause misfunction of the protein and genetic diseases. The analysis of micrographs can help to understand the distribution of proteins in the cell. ER protein's images analysis leads the researches to know that the same protein localization means the proteins have same function, and to understand the phylogenetic relationship by image's morphologies. This research uses SDA to find the best combination of features from original, skeletonized and brighter area images, which then utilized in SVM classification. The accuracy of this system can achieve 74% for the images acquired by the same hardware equipment.