Multiplex protein pattern unmixing using a non-linear variable-weighted support vector machine as optimized by a particle swarm optimization algorithm

• Non-linear machine learning is proposed for the first time for protein pattern unmixing.
• VW-SVM as optimized by PSO allows automatically flexible pattern feature extraction.
• Greatly improved performance is obtained via using VW-SVM for pattern unmixing.
• Results reveal non-linear VW-SVM has a great potential in protein pattern unmixing.

Most of the proteins locate more than one organelle in a cell. Unmixing the localization patterns of proteins is critical for understanding the protein functions and other vital cellular processes. Herein, non-linear machine learning technique is proposed for the first time upon protein pattern unmixing. Variable-weighted support vector machine (VW-SVM) is a demonstrated robust modeling technique with flexible and rational variable selection. As optimized by a global stochastic optimization technique, particle swarm optimization (PSO) algorithm, it makes VW-SVM to be an adaptive parameter-free method for automated unmixing of protein subcellular patterns. Results obtained by pattern unmixing of a set of fluorescence microscope images of cells indicate VW-SVM as optimized by PSO is able to extract useful pattern features by optimally rescaling each variable for non-linear SVM modeling, consequently leading to improved performances in multiplex protein pattern unmixing compared with conventional SVM and other exiting pattern unmixing methods.

Fundamental (mitochondrial or lysosomal) and mixed location patterns were both employed to compose training set. Their images sets were transformed into hybrid feature matrix by the object-based image model using individual learning strategy (ILS) and collective learning strategy (CLS). Utilizing variable-weighted support vector machine established a model between hybrid feature matrix and expected pattern fraction to fully interpreting the variation characteristic of protein dynamics. An unknown location pattern (fundamental or mixed) was applied to test the performance of established model and mapped into a set of mixture coefficients automatically.Figure optionsDownload as PowerPoint slide