Table 7. Structural profiles provide precursor specific signatures universally.
TP | FN | TN | FP | Sn (%) | Sp (%) | Acc(%) | |
Cow Data | |||||||
Classification with matrix | 42 | 0 | 36 | 6 | 100 | 85.71 | 92.85 |
Classification withoutmatrix | 37 | 5 | 34 | 8 | 88.09 | 80.95 | 84.52 |
Dog data | |||||||
Classification with matrix | 54 | 7 | 56 | 5 | 88.52 | 91.80 | 90.16 |
Classification withoutmatrix | 50 | 11 | 57 | 4 | 81.96 | 93.44 | 87.04 |
C. elegans Data | |||||||
Classification with matrix | 94 | 5 | 93 | 9 | 94.94 | 93.00 | 93.96 |
Classification withoutmatrix | 91 | 8 | 89 | 9 | 91.91 | 89.00 | 90.45 |
For never seen before datasets from cow, rat and C. elegans, the classification systems with the profiles developed from human specific precursors performed significantly better than the classification systems without human specific profiles. This suggests that the profiles have domain specific memory instead of instance based individual memory, making them useful for universal characterization of the precursors.