![]() This led to the classification of images into different classes based on local or global features. Research projects have focused on edge-based features (Wilson, 2002 Bern et al., 2004 ), texture analysis (Cumbaa & Jurisica, 2010 Ng et al., 2014 ) and spectral methods (Walker et al., 2007 ). Starting with edge detection in 1991 from robotically imaged crystallization trials (Ward et al., 1988 Zuk & Ward, 1991 ), crystallization image analysis has been subject to research. To overcome the obstacle of crystallization itself, automatic high-throughput crystallization robots were introduced (Thielmann et al., 2012 ) but the inspection of thousands of crystallization images still remains a burden. However, despite methodological advances (Birch et al., 2018 ), crystallization and the phasing problem remain the bottlenecks of macromolecular X-ray crystallography. Interestingly, purification conditions optimized for cryo-EM structure determination often show initial crystals in crystallization attempts (Stark & Chari, 2016 Chari et al., 2015 ). The resolution revolution in cryogenic electron microscopy (cryo-EM) accelerated the determination of molecular structures (Kühlbrandt, 2014 ) however, single-particle cryo-EM is fast becoming a rival technique. X-ray crystallography is the traditional method to determine atomic structures of macromolecules and is still the dominant technique based on the number of PDB (Protein Data Bank) entries in 2020 (RCSB Protein Data Bank, 2021 ). In addition, regions of droplets with the highest scoring probability found by the system are also available as images. These are immediately visible as colored frames around each crystallization well image of the inspection program. The outcome of the program is redistributed into the database as automatic real-time scores (ARTscore). To avoid high workloads for the control computer of the CrystalMation system, the computing is distributed over several workstations, participating voluntarily, by the grid programming system from the Berkeley Open Infrastructure for Network Computing (BOINC). With these two extremes it was found that an image processing rate of at least two times, but up to 58 times in the worst case, would be needed to reach the maximum imaging rate according to the deep learning network architecture employed for real-time classification. Two assumptions were made about the imaging rate. In detecting crystals AlexNet accomplished a better result, but with a lower threshold the mean value for crystal detection was improved for SqueezeNet. Four network architectures were compared and the SqueezeNet architecture performed best. Since the success rate of such a system is able to catch up with manual inspection by trained persons, it will become an important tool for crystallographers working on biological samples. The program uses manually scored crystallization trials deposited in a database of an in-house crystallization robot as a training set. To avoid the time-consuming and often monotonous task of manual inspection of crystallization plates, a Python-based program to automatically detect crystals in crystallization wells employing deep learning techniques was developed.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |