Table 1 The summarization of the related works

DeepDSC [6]

Used drug SMILES to extract drug features

Gene expression data is used to create a feature vector for the cell line

Processes the concatenated vector of drug and cell line features through three fully connected layers

Loewe synergy score

AuDNNsynergy [7]

Used drug SMILES to extract drug features

Gene expression, copy number, and genetic mutation data are used to create a comprehensive feature vector

Processes the drug and cell line features by passing them through fully connected layers

Loewe synergy score

SynPred [9]

Used drug SMILES to extract drug features

Utilized gene expression, copy number variation, methylation, global chromatin profiling, metabolomics, microRNA, and proteomics features

Utilized a fully connected subnetwork to integrate cell line features with the features of the two drugs

Loewe synergy, Bliss synergy, highest single agent model, and Zero interaction potency model

MatchMaker [8]

Used drug SMILES to extract drug features

Gene expression data is used to create a feature vector for the cell line

Concatenate each drug feature with the cell line feature and pass the combined data through two parallel fully connected subnetworks. The outputs of these subnetworks are then concatenated and fed into a final fully connected subnetwork

Sensitivity and synergy scores

TranSynergy [11]

Drug features extracted from 2041 selected genes by drug-target interaction

Cell line features extracted from 2041 gene dependency or gene expression

Used a transformer deep learning model to map concatenated drug and cell line features

Loewe synergy score

DeepDDS [12]

Convert the drug SMILES to a graph network

Used gene expression of cell line as a feature vector

Used a graph attention network for drugs and fully connected subnetwork for cell lines then concatenate them and fed to other fully connected layers

Binary classification of drugs (synergistic or antagonistic)