University of California San Diego researchers say they have developed a "visible" neural network and used it to construct DCell, a virtual model of a functioning brewer's yeast cell.
They accomplished this by accumulating a central dataset of cell biology knowledge and creating a hierarchy of cellular components, and then mapping standard machine-learning algorithms to this dataset.
DCell's learning is guided only by real-world cellular behaviors and limitations coded from about 2,500 known cellular components. The researchers feed information about genes and genetic mutation to DCell, which predicts cellular behaviors.
The team trained DCell on several million genotypes and determined it could model cellular growth with almost as much accuracy as an actual, lab-cultured cell.
The researchers are working to generate experimental data to build DCell for human cancer, in order to determine the best way to tailor this virtual cell for a patient's unique biology.
From UC San Diego Health System
View Full Article
Abstracts Copyright © 2018 Information Inc., Bethesda, Maryland, USA
No entries found