Table 1. Summary of feature columns and fragment sizes.

Feature Name	No. of columns	No. of residues per fragment window
21-bit Sparse Coding	21 per residue	11 sequential residues (a sliding window of size 5)
GAC	20 per fragment	whole protein sequence
PSSM	20 per residue	11 sequential residues
EC	1 per residue	11 sequential residues
LN	5 per residue	11 sequential residues
Normalized ASAs	3 per residue	11 spatial residues (a neighboring window of size 10)
Physicochemical (PC) property	10 per fragment	21 spatial residues (a neighboring window of size 20)
Predicted secondary structure	8 per residue	11 sequential residues

Since the GAC is calculated from a single protein sequence, for each coding fragment, a GAC vector will be appended. For the PC feature, for a coding fragment a list of 21 neighboring residues will return 10 values.