Quad-PRE: A Hybrid Method to Predict Protein Quaternary Structure Attributes

Table 2

Summary of the considered features, where 𝑦 denotes one of the three secondary structure states and 𝑥 denotes one of the 20 common AAs.

Feature sets	Description
Sequence-based (79)	Sequence length (1); Composition vector (20); The number of AAs in the sequence belonging to {R group, Electronic group, Hydrophobicity group, Exchange group} (18); First and second order composition moment vector (40)
PSSM-based (203)	From the PSSM matrix
Secondary structure (217)	Based on the features utilized in the PSI-Pred method (90); Based on the predicted secondary structure which describes collocation; of helical and strand segments (127)
Average RSA based (23)	Average RSA of the residues with AA type 𝑥 (20); Average RSA of the residues with secondary structure type 𝑦 (3)
Average isoelectric point (1)	𝑝𝐼=1/𝑁∑^𝑁𝑖=1𝑝𝐼𝑖, the 𝑝𝐼𝑖 values in the paper [11]
Auto-correlation functions based on FH𝑖, EH𝑖, and Hp indices (25)	𝐴^𝑎𝑛=1/(𝑁−𝑛)∑^𝑁−𝑛𝑖=1𝑎𝑖𝑎𝑖+𝑛, where 𝑎 defines the corresponding physicochemical properties, such as two hydrophobicity indices (the Fauchere-Pliska’s (FH) with 𝑛=1,2,…,10 and the Eisenberg’s (EH) 𝑛=1,2,…,6), and hydropathy (HP) index with 𝑛=1,2,…,9.
Auto-correlation functions based on cumulative FH𝑖 index (6)	𝐴^𝑎𝑛=∑^𝑁−𝑛𝑖=1∑^𝑖𝑗=1𝑎𝑗×∑^𝑖+𝑛𝑗=1𝑎𝑗/(𝑁−𝑛), where 𝑎 is the FH index with 𝑛=1,2,…,6.
Sum of hydrophobicities based on FH𝑖 and EH𝑖 (2)	𝐻^𝑎sum=∑^𝑁𝑖=1𝑎𝑖, where 𝑎 is the FH or the EH index.
R groups (5)	RG𝑖, where 𝑖=1 corresponds to nonpolar aliphatic AAs (AVLIMG), 𝑖=2 to polar uncharged AAs (SPTCNQ), 𝑖=3 to positively charged AAs (KHR), 𝑖=4 to negative AAs (DE), and 𝑖=5 to aromatic AAs (FYW); the composition percentage of each group in the sequence is computed
Electronic groups (5)	EG𝑖, where 𝑖=1corresponds to electron donor AAs (DEPA), 𝑖=2 to weak electron donor AAs (LIV), 𝑖=3 to electron acceptor AAs (KNR), 𝑖=4 to weak electron acceptor AAs (FYMTQ), and 𝑖=5 to neutral AAs (GHWS); the composition percentage of each group in the sequence is computed
Blast based (30)	Refer to subsection “Features”
GLAM2-based (30)	Refer to subsection “Features”
GIBBS-based (6)	Refer to subsection “Features”