GEFA: Early Fusion Approach in Drug-Target Affinity Prediction

Nguyen, Tri Khoa; Nguyen, Thin; Le, Thao Minh; Tran, Truyen

doi:10.1109/tcbb.2021.3094217

Cited by 55 publications

(64 citation statements)

References 66 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To balance between 3D structural information and simplicity, 2D representation via attributed graph can be used. Previous works (Nguyen et al, 2021;Jiang et al, 2020) have been using protein structure graph representation for DTA prediction. The contact/distance map is used as the adjacency matrix of an attributed graph where each node represents a residue and edge represents the contact/distance between residues.…”

Section: Protein Graph Representationmentioning

confidence: 99%

“…The contact/distance map is used as the adjacency matrix of an attributed graph where each node represents a residue and edge represents the contact/distance between residues. The node attribute can be simply a one-hot encoding of residue type (Jiang et al, 2020) or an embedding vector of the residue obtained from the language model (Nguyen et al, 2021).…”

Section: Protein Graph Representationmentioning

confidence: 99%

“…Over the years, many machine learning-based DTA prediction methods (Cichonska et al, 2017;Öztürk et al, 2018;Nguyen et al, 2020Nguyen et al, , 2021 have been proposed. However, these computational methods face the coldstart challenge where the model performance drops in novel drugs or targets, which are common in drug discovery or drug repurposing.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Mitigating cold start problems in drug-target affinity prediction with interaction knowledge transferring

Nguyen¹,

Nguyen²,

Tran³

2022

Preprint

Self Cite

View full text Add to dashboard Cite

Motivation: Predicting the drug-target interaction is crucial for drug discovery as well as drug repurposing. Machine learning is commonly used in drug-target affinity (DTA) problem. However, machine learning model faces the cold-start problem where the model performance drops when predicting the interaction of a novel drug or target. Previous works try to solve the cold start problem by learning the drug or target representation using unsupervised learning. While the drug or target representation can be learned in an unsupervised manner, it still lacks the interaction information, which is critical in drug-target interaction. Results: To incorporate the interaction information into the drug and protein interaction, we proposed using transfer learning from chemical-chemical interaction (CCI) and protein-protein interaction (PPI) task to drug-target interaction task. The representation learned by CCI and PPI tasks can be transferred smoothly to the DTA task due to the similar nature of the tasks. The result on the drug-target affinity datasets shows that our proposed method has advantages compared to other pretraining methods in the DTA task.

show abstract

Section: Protein Graph Representationmentioning

confidence: 99%

Section: Protein Graph Representationmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Mitigating cold start problems in drug-target affinity prediction with interaction knowledge transferring

Nguyen¹,

Nguyen²,

Tran³

2022

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…Their approach outperforms 3D-CNNs on virtual screening, while also producing interpretable results that indicate the contributions of each ligand atom and pocket residue to the classification decision. DGraphDTA [ 50 ] and GEFA [ 52 ] are methods that improve upon the earlier GraphDTA [ 76 ], which combines an atom-based graph representation for the ligand and a sequence representation for the protein. DGraphDTA first predicts the protein contact map based on the protein’s amino acid sequence, and then it constructs a protein graph given the contact map.…”

Section: Protein–ligand Interactionmentioning

confidence: 99%

“…As the number of datasets and studies that GNN-based methods are tested is still low, it is unknown how well the methods discussed above generalize. Moreover, in the case where the protein–ligand complex is not given (Figure 3B ), there is no sufficient evidence that the contributions of the ligand atoms presented in [ 49 ] or the protein–ligand connections learned in [ 52 ] correlate with actual structural protein–ligand data. However, it is evident that — whether there is enough protein–ligand complex data or not-using a graph representation and applying GNN-based architectures can improve binding affinity predictions compared with CNN-like approaches and sequence-based models [ 46 , 50 ].…”

Section: Protein–ligand Interactionmentioning

confidence: 99%

Graph representation learning for structural proteomics

Fasoulis

Παλιούρας

Kavraki

2021

Emerging Topics in Life Sciences

View full text Add to dashboard Cite

The field of structural proteomics, which is focused on studying the structure–function relationship of proteins and protein complexes, is experiencing rapid growth. Since the early 2000s, structural databases such as the Protein Data Bank are storing increasing amounts of protein structural data, in addition to modeled structures becoming increasingly available. This, combined with the recent advances in graph-based machine-learning models, enables the use of protein structural data in predictive models, with the goal of creating tools that will advance our understanding of protein function. Similar to using graph learning tools to molecular graphs, which currently undergo rapid development, there is also an increasing trend in using graph learning approaches on protein structures. In this short review paper, we survey studies that use graph learning techniques on proteins, and examine their successes and shortcomings, while also discussing future directions.

show abstract