Skip to main content
Fig. 7 | BMC Genomics

Fig. 7

From: DCSE:Double-Channel-Siamese-Ensemble model for protein protein interaction prediction

Fig. 7

DCSE consists of three main parts, which are the embedding layer, the feature extraction layer, the prediction layer. The input of our model is protein sequence1 and protein sequence2. Both the sequence is truncated the first 1000 amino acids if the sequence is more than 1000 or padded to 1000 if the sequence is less than 1000. In the embedding layer, each amino acid is mapped into a 25-dimensional vector so each protein can be represented by a 1000*25 feature matrix(S1, S2). Both the S1 and S2 are then fed into the siamese-based feature extraction layer. Then the output of the feature extraction layer is concatenated together and put into the prediction layer which finally gives the predictive result of our model

Back to article page