Open AccessApplication of a neural network in high‐throughput protein crystallography
Author(s) -
Berntson A.,
Stojanoff V.,
Takai H.
Publication year - 2003
Publication title -
journal of synchrotron radiation
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.172
H-Index - 99
ISSN - 1600-5775
DOI - 10.1107/s0909049503020855
Subject(s) - throughput , artificial neural network , computer science , automation , protein crystallization , set (abstract data type) , diffraction , quality (philosophy) , convergence (economics) , test set , artificial intelligence , pattern recognition (psychology) , algorithm , chemistry , engineering , optics , physics , mechanical engineering , telecommunications , organic chemistry , crystallization , economics , wireless , programming language , economic growth , quantum mechanics
High‐throughput protein crystallography requires the automation of multiple steps used in the protein structure determination. One crucial step is to find and monitor the crystal quality on the basis of its diffraction pattern. It is often time‐consuming to scan protein crystals when selecting a good candidate for exposure. The use of neural networks for this purpose is explored. A dynamic neural network algorithm to achieve a fast convergence and high‐speed image recognition has been developed. On the test set a 96% success rate in identifying properly the quality of the crystal has been achieved.