Skip to main content
Fig. 6 | BMC Bioinformatics

Fig. 6

From: DeepDist: real-value inter-residue distance prediction with deep residual convolutional network

Fig. 6

The overall workflow of DeepDist for both real-value distance map prediction and multi-class distance map prediction. Given a sequence, DeepAln and DeepMSA are called to search it against sequence databases to generate two kinds of multiple sequence alignments (MSAs), which are used to generate four sets of features (COV, PLM, PRE, OTHER), respectively. The four sets of features are used by four deep networks (COV Net, PLM Net, PRE Net, and OTHER Net) to predict both real-value distance (real-dist) map and multi-class distance (multi-class) map, respectively. The real-value distance maps (or multi-class distance maps) of the individual networks are averaged to produce the final real-value distance map (or multi-class distance map)

Back to article page