FIG 4.

(a) Schematic representation of the four input representations tested. As demonstrated above, a thymine-to-cytosine mutation in the DNA sequence, which corresponds to a cysteine-to-arginine amino acid mutation in the protein sequence, was represented in four different ways. The binary representation method marked this as position 1, since a mutation occurred in that position. The scored representation method assigned a score of −3 to the amino acid mutation, based on the Blosum 62 matrix. The amino acid and nucleotide representation methods represented the mutation using the sparse coding, where all the possible features were represented, the corresponding feature to the mutation was marked 1; the remaining features were marked −1. *, additional features from PointFinder and insertion and deletion options. (b) An example of scored and nucleotide representation methods. Pos, position.