. 2017 Apr 26;19(6):1183–1202. doi: 10.1093/bib/bbx041

Table 1.

Overview of the compared methods and their respective class

Class

Method

Method or formula

Supervised (SVM)

BLM_dt [37]

SVM+chemical and sequence similarity

BLM_d [37]

SVM+drug chemical similarity

BLM_t [37]

SVM+target sequence similarity

Supervised (GR+MF)

GRMF_dt [14]

GR+MF+chemical and sequence similarity

GRMF_d [14]

GR+MF+chemical and sequence similarity

GRMF_t [14]

GR+MF+chemical and sequence similarity

Supervised (WKNKN+GR+MF)

wGRMF_dt [14]

WKNKN+GR+MF+chemical and sequence similarity

wGRMF_d [14]

WKNKN+GR+MF+chemical and sequence similarity

wGRMF_t [14]

WKNKN+GR+MF+chemical and sequence similarity

Unsupervised (MF)

MF [14]

MFm

MF mean score CV

MFb

MF best score CV

MFw

MF weight score CV

Unsupervised (Projection)

BPR [40]

Random walk

NBI [41]

Jac [40]

Jac similarity

Euc [40]

Euclidean distance

Cos [40]

Cosine similarity

Pea [40]

Pearson correlation

Unsupervised (LCP)

CAR [30]

C A R (d_{x}, t_{y}) = C N (d_{x}, t_{y}) \times L C L (d_{x}, t_{y})

CJC [30]

C J C (d_{x}, t_{y}) = \frac{C A R (d_{x}, t_{y})}{| N (d_{x}) \cup N (t_{y}) |}

CPA [30]

C P A (d_{x}, t_{y}) = (e N (d_{x}) + C A R (d_{x}, t_{y})) \times (e N (t_{y}) + C A R (d_{x}, t_{y}))

CAA [30]

C A A (d_{x}, t_{y}) = \sum_{s \in C N (d_{x}, t_{y})} \frac{| γ (s) |}{{log}_{2} | N (s) |}

CRA [30]

C R A (d_{x}, t_{y}) = \sum_{s \in C N (d_{x}, t_{y})} \frac{| γ (s) |}{| N (s) |}

Note. N(d_x) indicates the first neighbours of the drug d_x. CN(d_x; t_y) are the CNs of drug d_x and target t_y, respectively, and LCL(d_x; t_y) are their local community links. In respect to the (d_x; t_y) link, eN(d_x) and eN(t_y) represent the neighbours of drug d_x and target t_y, respectively, which do not belong to the set of CN(d_x; t_y), thus indicated as external neighbours. γ(s) are the neighbours of s, which are also CN(d_x; t_y), while N(s) are all the neighbours of s. SVM: Support Vector Machine; GR: Graph Regularization; MF: Matrix Factorization; MFb: best Matrix Factorization; MFm: mean Matrix Factorization; MFw: weighted Matrix Factorization; WKNKN: Weighted K Nearest Known Neighbours; BLM: Bipartite Local Model; GRMF: Graph Regularized Matrix Factorization; BPR: Bipartite Projection via Random-walk; NBI: Network-Based Inference; Jac: Jaccard; Euc: Euclidean; Cos: Cosine; Pea: Pearson; CAR: Cannistraci-Alanis-Ravasi; CJC: Cannistraci-Jaccard; CPA: Cannistraci-Preferential-Attachment; CAA: Cannistraci-Adamic-Adar; CRA: Cannistraci-Resource-Allocation.