-This SVM was trained using LIBSVM-3.12, with the -b 1 option used to enable probability outputs.
-Features should be scaled before use using svm-scale (provided by LIBSVM-3.12) and the scaling parameters available at the SuSPect website.
-Prediction should be carried out using svm-predict -b 1
-Output will be a score from 0-1, which should be multiplied by 100 to give the SuSPect score.

The following features are included in SuSPect-All:

1 - PPI betweenness from STRING
2 - PPI closeness from STRING
3 - PPI coreness from STRING
4 - PPI degree from STRING
5 - DDI betweenness from DOMINE
6 - DDI closeness from DOMINE
7 - DDI coreness from DOMINE
8 - DDI degree from DOMINE
9 - Domain bigram betweenness from Pfam
10 - Domain bigram closeness from Pfam
11 - Domain bigram coreness from Pfam
12 - Domain bigram degree from Pfam
17 - Protein's aliphatic index
18 - Protein's GRAVY (grand average of hydropathy)
24 - Number of annotations for that position in the UniProt FT table
25 - BLOSUM score for the SAV
26 - Charge change caused by the SAV
27-31 - Changes in principle components of Amino Acid index
32 - Wild-type amino acid is Pro
33 - Variant amino acid is Pro
34 - Wild-type amino acid is Gly
35 - Variant amino acid is Gly
36 - PSSM score of wild-type
37 - PSSM score of variant
38 - Feature 36-Feature 37
39 - PSSM Info score for the column
40 - Pfam HMM emission score for the wild-type amino acid
41 - Pfam HMM emission score for the variant amino acid
42 - Feature 40-Feature41
43 - Jensen-Shannon divergence
44 - Percentage sequence identity with the closest homologue to lack the wild-type amino acid at that position
45 - Percentage sequence identity with the closest homologue to have the variant amino acid at that position
46 - BLAST E-value with the closest homologue to lack the wild-type amino acid at that position
47 - BLAST E-value with the closest homologue to have the variant amino acid at that position
48 - Proportion of gaps in the column of the MSA
51 - Structural information from PDB
52 - Structural information from Phyre2
53 - No structural information available
54 - Secondary structure is G/H/I (NACCESS, 0/1)
55 - Secondary structure is B/E (0/1)
56 - Secondary structure is S/T (0/1)
57-60 - Bond energy from H-bonding for N1, N2, O1, O2 (NACCESS)
61-3 - Probability of alpha/beta/coil from NetSurfP
64 - IUPred score for the wild-type sequence
65 - IUPred score for the variant sequence
66 - Feature 64-Feature 65
67 - Phi backbone angle
68 - Psi backbone angle
69 - ANCHOR probability for the wild-type sequence
70 - ANCHOR class for the wild-type sequence (0/1)
71 - ANCHOR probability for the variant sequence
72 - ANCHOR class for the variant sequence (0/1)
73 - Feature 69-Feature 71
74 - Feature 70-Feature 72
76 - Labelled as interacting in PISite (0/1)
78 - Betweenness centrality in a residue interaction network (normalised as a z-score)
79 - In a surface pocket according to fpocket (0/1)
80 - Predicted relative solvent accessibility from NetSurfP
81 - All-atom RSA from NACCESS
82 - Sidechain RSA from NACCESS
83 - Mainchain RSA from NACCESS
84 - Non-Polar RSA from NACCESS
85 - Polar RSA from NACCESS
86 - Labelled in the Catalytic Site Atlas (0/1)
87 - Crystallographic B factor
88 - Labelled as TRANSMEM in UniProt KB (0/1)
89 - Labelled as interacting in ProtInDB (0/1)