Automatic classification of sub-microlitre protein-crystallization trials in 1536-well plates

Abstract

A technique for automatically evaluating microbatch (400 nl) protein-crystallization trials is described. This method addresses analysis problems introduced at the sub-microlitre scale, including non-uniform lighting and irregular droplet boundaries. The droplet is segmented from the well using a loopy probabilistic graphical model with a two-layered grid topology. A vector of 23 features is extracted from the droplet image using the Radon transform for straight-edge features and a bank of correlation filters for microcrystalline features. Image classification is achieved by linear discriminant analysis of its feature vector. The results of the automatic method are compared with those of a human expert on 32 1536-well plates. Using the human-labeled images as ground truth, this method classifies images with 85% accuracy and a ROC score of 0.84. This result compares well with the experimental repeatability rate, assessed at 87%. Images falsely classified as crystal-positive variously contain speckled precipitate resembling microcrystals, skin effects or genuine crystals falsely labeled by the human expert. Many images falsely classified as crystal-negative variously contain very fine crystal features or dendrites lacking straight edges. Characterization of these misclassifications suggests directions for improving the method.

Automated High Throughput Protein Crystallization Screening at Nanoliter Scale and Protein Structural Study on Lactate Dehydrogenase

“…Those differences have profound implications on protein growth methods, x-ray diffraction data collection strategies and data qualities [15][16][17][18][19][20][21][22][23][24],…”

“…The interactions among protein molecules are also weak. Therefore, protein crystals are generally less ordered and show lower diffraction qualities compared with traditional crystals [15][16][17][18][19][20][21][22][23][24][25][26][27][28][29][30][31][32][33][34],…”

“…Some studies show protein growth mechanisms are similar to those for small molecule crystals, which include spiral dislocations, two-dimensional nucleation on surfaces, random nucleation and normal growth [19,21,35], It has been shown by AFM study that the characteristic kinetic parameters of protein crystal growth are significantly lower than those for traditional crystals by a factor of two or three magnitudes [15,17,[36][37][38][39][40][41], In practice, small molecule crystals can grow in minutes or days while protein crystallization takes days to weeks or even months.…”

SPINE high-throughput crystallization, crystal imaging and recognition techniques: current state, performance analysis, new technologies and future aspects

“…Several systems for analyzing and classifying images are being developed (see, for example, Bern et al, 2004;Cumbaa et al, 2003;Spraggon et al, 2002;Wilson, 2004). Whilst the fundamental aim is the detection of crystals, reliable classification of other outcomes potentially provides vital feedback to aid the development of optimization protocols.…”

Image-Based Classification for Automating Protein Crystal Identification

“…Hough Transform is utilized to detect the straight lines. Two values are taken as the geometry features, as ever mentioned by Cumbaa et al [5], the total length and the maximum length of the lines detected in the image.…”