Fig. 10: Setup of NetTCR model. | Communications Biology

Fig. 10: Setup of NetTCR model.

From: NetTCR-2.0 enables accurate prediction of TCR-peptide binding by using paired TCRα and β sequence data

Fig. 10

The CDR3 and peptide sequences are encoded using the BLOSUM50 matrix. The encoded sequences are passed independently through a 1D convolutional layer and a max-pooling layer. The convolutional filter size is set to {1, 3, 5, 7, 9}, and for each filter size, 16 filters are used. The extracted features are then concatenated and fed into a dense layer with 32 hidden units. The output of the network consists of a single neuron, giving the binding probability.

Back to article page