Fig. 1
From: Freeprotmap: waiting-free prediction method for protein distance map
Illustration of our proposed framework, which consists of three modules: representation generation module, representation optimization module, and regression prediction module. In the representation generation module, ESM-2 generates attention maps of the input protein, which are then downscaled through group pooling to form the input representation (input feature maps). In the representation optimization module, the representation is further optimized by the R-former. The regression prediction module has two branches: one for distance map prediction and one for error prediction