(a) Input example sequence along with PSI-BLAST profile matrix of dimensions n × 20, with a residue circled to show the encoding steps. (b) Example wmer with w = 3 giving length seven with extracted features from the original PSI-BLAST matrix. (c) Encoded vector of length 7 × 20 formed by linearizing the sub-matrix. (d) Flexible encoding showing three residues in the center using the finer representation, and two residues flanking the central residues on both sides using a coarser representation as an averaging statistic. Length of this vector equals 5 × 20.
Rangwala et al. BMC Bioinformatics 2009 10:439 doi:10.1186/1471-2105-10-439