Computational Systems Biology Methods and Protocols.7z

2.2.9 Solvent Accessible
Surface Area

The solvent accessible surface area was predicted using RVP-Net

[40, 41], which shows the proportion of solvent accessible area for

each amino acid in a protein sequence. In PGlcS, the RVP-Net

outputs were normalized to be approximately 0–1 for every amino

acid close to the O-GlcNAcylation sites [10].

2.2.10 First- and
Second-Order Composition
Moment Vector (CMV)

Two encoding methods were used to obtain reduced amino acid

sequences according to the physicochemical properties of the

amino acids. Based on their acid–base properties, the 20 amino

acid residues were classified into three groups: acidic (AAaci), {D,E};

basic (AAbas), {K,H,R}; and neutral (AAneu), {A,C,F,G,I,L,M,N,P,

Q,S,T,V,W,Y} amino acids. And based on their hydrophobicity, the

20 amino acid residues were classified into three other groups:

internal (AAint), {F,I,L,M,V}; external (AAext), {D,E,H,K,N,Q,

R}; and ambivalent (AAamb), {A,C,G,P,S,T,W,Y} amino acids.

To reflect the order of acidic and hydrophobic amino acids in

the sequences surrounding candidate O-GlcNAcylation sites,

CMVs were defined as

CMVik¼

Pxi

j¼ 1

nijk

∏kd¼ 0 ðÞNd

wherek¼0, 1 is the order of CMV;Lis the sequence length;iis

theith amino acid with acid–base {AAaci,AAbas,AAneu} and hydro-

phobicity {AAintl,AAext,AAamb} properties;xjis the total number

ofiresidues in the corresponding reduced amino acid sequence;

andnijis thejth position ofiresidues in the corresponding reduced

sequence. Whenk¼0, the CMV reduces to the content of theith

amino acid. Therefore, Acontent/Hcontent (the first order of

CMV) and ACMV/HCMV (the second order of CMV) were

defined to reflect the sequence order of acidic and hydrophobic

amino acids, respectively [8].

2.3 Assessing
Predictive
Performance

The predictive performances of the O-GlcNAcylation sites predic-

tors were assessed by calculating four commonly used measure-

ments, sensitivity (Sn), specificity (Sp), accuracy (Acc), and

Matthew’s correlation coefficient (MCC), as follows:

Sn¼

TP

TPþFN

Sp¼

TN

TNþFP

Acc¼

TPþTN

TPþTNþFPþFN

242 Cangzhi Jia and Yun Zuo