Feature analysis for classification of trace fluorescent labeled protein crystallization images

Abstract

BackgroundLarge number of features are extracted from protein crystallization trial images to improve the accuracy of classifiers for predicting the presence of crystals or phases of the crystallization process. The excessive number of features and computationally intensive image processing methods to extract these features make utilization of automated classification tools on stand-alone computing systems inconvenient due to the required time to complete the classification tasks. Combinations of image feature sets, feature reduction and classification techniques for crystallization images benefiting from trace fluorescence labeling are investigated.ResultsFeatures are categorized into intensity, graph, histogram, texture, shape adaptive, and region features (using binarized images generated by Otsu’s, green percentile, and morphological thresholding). The effects of normalization, feature reduction with principle components analysis (PCA), and feature selection using random forest classifier are also analyzed. The time required to extract feature categories is computed and an estimated time of extraction is provided for feature category combinations. We have conducted around 8624 experiments (different combinations of feature categories, binarization methods, feature reduction/selection, normalization, and crystal categories). The best experimental results are obtained using combinations of intensity features, region features using Otsu’s thresholding, region features using green percentile G 90 thresholding, region features using green percentile G 99 thresholding, graph features, and histogram features. Using this feature set combination, 96% accuracy (without misclassifying crystals as non-crystals) was achieved for the first level of classification to determine presence of crystals. Since missing a crystal is not desired, our algorithm is adjusted to achieve a high sensitivity rate. In the second level classification, 74.2% accuracy for (5-class) crystal sub-category classification. Best classification rates were achieved using random forest classifier.ContributionsThe feature extraction and classification could be completed in about 2 s per image on a stand-alone computing system, which is suitable for real time analysis. These results enable research groups to select features according to their hardware setups for real-time analysis.

Time-Dependent Multi-Light-Source Image Classification Combined With Automated Multidimensional Protein Phase Diagram Construction for Protein Phase Behavior Analysis

“…Balanced accuracy (the average of recall and precision) was used as an evaluation parameter because protein phase diagram image data sets often deal with a class imbalance. 18,34 This class imbalance is not only an aspect to take into account during model training via proper class representation, but also during model evaluation. Six feature sets have been evaluated in this study: performed.…”

“…The work of Bruno et al 17 uses similar classes ("clear," "crystal," "precipitate," "other"), whereas Sigdel et al and Cumbaa et al used a 3-class system ("clear," "crystal," "other"). 18,21 In addition to the classes, the type of classifier was taken into account. As listed in Table 2, deep convolutional neural networks (CNNs) were used by Bruno et al, whereas a random forest classifier was used by the Sigdel et al and Cumbaa et al The type of classifier is of interest because of the required computational time and expertise to design and train a classification model.…”

“…Bruno et al 17 Cumbaa et al 21 Sigdel et al 18 This The values are given in mean ± standard deviation, which are determined based on the corresponding evaluation parameters for all considered classes. misclassified crystallized formulation in plate B (C9, C12, D4) and plate C (A1), which also contains a few small crystals, it can be concluded that the classifier does not perform well for the classification of a few small crystals.…”

“…One of the major issues in protein phase behavior image classification is the number of protein phase behavior types, subtypes, and the coexistence of these (sub)types. 18 For example, crystallization can be in the form of needle crystals or three-dimensional crystals. In addition, these subtypes can coexist in a single formulation or coexist with another protein phase behavior type, such as precipitates.…”

“…A strategy to improve the accuracy of classification models is the use of different data sources, which have been explored to find more distinctions between protein phase behavior (sub)types. Evaluated sources include, but are not limited to, protein trace-labeling in combination with green fluorescence, 18,23,24 second-order nonlinear imaging of chiral crystals, 25 ultraviolet (UV) light, 26,27 and two-photon excited UV fluorescence. 28 Despite these alternatives, the main data source remains images obtained with visible light.…”