Sphatika is a crystal image classification tool for high throughput X-ray crystallography. Sphatika classifies protein crystals into two broad categories, one comprising crystal hits and harvestable crystals and the other comprising empty wells, clear drops and precipitates. The accuracy achieved is more than 92% on harvestable crystals and crystal hits together with a 20% false positive rate. With Sphatika running on a cluster, hundreds of thousands of images can be processed in a day, and the results of classification can be directly used, with minimal human intervention.

  Key Features

• Advanced and proprietary image processing techniques used to extract significant features
• Accuracy north of 94% on several data sets and north of 90% on the whole
• Scalable on industry standard hardware clusters
• Automated two way (Crystal/Non-crystal) or four way (Crystal/ Harvestables/ Clear/ Precipitate) classifier.

  Architecture