Fast protein structure comparison through effective representation learning with contrastive graph neural networks
Fig 3
The contrastive learning framework for protein structure representation learning.
At each iteration, raw features Xq and Xk are extracted from the query protein structure and the key protein structure, respectively. Then, descriptors yq and yk are encoded by GNN encoder and
, respectively. The value of loss function guides the optimization of the parameters θq of
while the parameters θk are updated based on θq. At the end of the current iteration, yk will enqueue as a negative sample for the next iteration.