Alphabet reduction and distributed vector representation based method for classification of antimicrobial peptides

Hydropathy and Conformational Similarity-Based Distributed Representation of Protein Sequences for Properties Prediction

“…Comparison of best result of current study and results of a previous study[20] the reduced alphabet based representation provides additional information about the structural or physicochemical changes which cannot be captured by ProtVec alone. The improved performance is quite evident in the two binding pocket tasks.…”

“…It must be mentioned here that Yang et al [20] employed the original twenty letter representation for embedding vectors in their ProtVec model and later trained the sequences for each task with the embedding vectors as input to a gaussian regression model and reported the results. We have illustrated our best results and results of Yang et al [20] in Table 6. In our work, we used an SVM regressor.…”

“…With this new reduced alphabet representation, they obtained considerable improvement in performance with and without synergistic combination with original ProtVec embeddings. Surana et al[20] applied a similar technique for the classification of antimicrobial peptides. The addition of reduced alphabet based embedded representation provides additional information in terms of the physicochemical properties.…”