PREDICTING 3D STRUCTURES OF LYSIN LOOP REGIONS FOR SYNTHESIS
Automated flow chemistry is not currently able to synthesize double domain proteins (lysins) due to loop regions of peptide fragments less than 10 amino acids that connect the domains. These regions are difficult to model because of the high dependency of the structure on external environments
Image created by student author,
AlphaFold
AlphaFold is a protein structure prediction system created by Google's DeepMind. It is built based on a convolutional neural network capable of creating highly accurate models of proteins.
WHAT ARE THE LIMITATIONS?
AlphaFold is currently unable to predict loop region structures due to the lack of hydrogen bonding, which the program relies on as a marker for structure prediction.
​
Our Innovation: AlphaFold.Loop
LyseDevice utilizes a refined algorithm, AlphaFold.Loop. The system will be optimized through stochastic gradient descent to predict the 3D structures of non-hydrogen bond lysin loop regions for synthesis by using different markers of prediction, focusing on domain end-to-end interactions.