research process

Psi-Blast

Gain/Losslist

Fold1

Border

Fold3

AAdatabase

Predictor

Train-set

base on sequence

Fold4

SS

SA

Disorder

Gain/Loss

5AAindex

Feature 1. Rank features by mRMR selection 2. Select features by increasing AUC based on SVM classifier

Scratch

Fold2

Disopred

Featurecaculation

PSSM

Output

Test-set

5-fold

Parameters SVM model training Libsvm (R package)

KNN

Psi-Pred

Fold5