Predictingprotein-protein interactions usingGCN-based encoder-decoder combined with a transformer

ODTÜ, Deniz Bilimleri Enstitüsü
Predicting protein-protein interactions using GCN-based encoder-decoder combined with a transformer Proteins are large and complex molecules that play countless roles throughout the biological world. They interact through their interfaces to fulfill essential functions in the cell. They bind to their partners in a particular manner and form complexes that highly affect understanding of the biological pathways in which they are involved. Any abnormal interactions may cause interruptions of their functions and may have detrimental effects on the organism. Furthermore, investigating residue interactions on protein-protein interfaces is essential for drug discovery in pharmaceutical research. As experimental data accumulates, artificial intelligence comes to the stage, and recent groundbreaking applications of AI profoundly impact the structural biology field. This talk will present our work on developing a graph-based framework for investigating protein-protein interfaces. Our framework converts protein-protein interfaces from PDB 3D coordinates files into graphs with nodes representing interface residues and edges corresponding to the residue interactions. Each residue is represented as a vector with the features: residue type, polarity, residue charge, relative accessible surface area (ASA), pair potentials, and backbone dihedral angles. Three edges are constructed: sequential edges, radius edges, and K-nearest neighbor edges. The graphs are further provided to a graph convolutional networkbased encoder-decoder architecture combined with a transformer to generate embeddings. To train and test the encoder-decoder part, we use a large set that contains more than 500.000 PPIs deposited from PDB. Then, these embeddings are fed to a Graph Interaction Network, which is composed of a succession of graph convolution layers (GCL), non-linear activation (ReLU), and pooling layers. For the downstream task, which is the scoring docking model, we use 271.830 interfaces which are split into training, validation, and test sets. Positive samples come from PDB and PIFACE, and the negative data comes from PPI4DOCK and DOCKGROUND. Our method achieved 0.91 accuracy for the test set and outperformed existing CNN models, which achieved around 0.75-0.81 accuracy for the test set and GNN architectures, such as DeepRank-GNN and GNN-DOVE, whose accuracies are 0.89 and 0.88 respectively.


Prediction of protein-protein interactions from sequence using evolutionary relations of proteins and species
Güney, Tacettin Doğacan; Can, Tolga; Department of Computer Engineering (2009)
Prediction of protein-protein interactions is an important part in understanding the biological processes in a living cell. There are completely sequenced organisms that do not yet have experimentally verified protein-protein interaction networks. For such organisms, we can not generally use a supervised method, where a portion of the protein-protein interaction network is used as training set. Furthermore, for newly-sequenced organisms, many other data sources, such as gene expression data and gene ontolog...
Prediction of enzyme classes in a hierarchical approach by using SPMap
Yaman, A.; Atalay, Mehmet Volkan; Atalay, Rengül (2010-04-01)
Enzymes are proteins that play important roles in biochemical reactions as catalysts. They are classified based on the reaction they catalyzed, in a hierarchical scheme by International Enzyme Commission (EC). This hierarchical scheme is expressed in four-level tree structure and a unique number is assigned to each enzyme class. There are six major classes at the top level according to the reaction they carried out and sub-classes at the lower levels are further specific reactions of these classes. The aim ...
Fast Screening of Protein-Protein Interactions Using Forster Resonance Energy Transfer (FRET-) Based Fluorescence Plate Reader Assay in Live Cells
Durhan, Seyda Tugce; Sezer, Enise Nalli; Son, Çağdaş Devrim; Küçük Baloğlu, Fatma (2022-11-01)
Protein-protein interactions (PPIs) have great importance for intracellular signal transduction and sustaining the homeostasis of an organism. Thus, the identification of PPIs is necessary to better understand the downstream signaling functions of the proteins in healthy and pathological conditions. Forster resonance energy transfer (FRET) between fluorescent proteins (FPs) is a powerful tool for detecting PPIs in living cells. In literature, FRET analysis methods such as donor photobleaching (FLIM), accept...
Predicting the effect of hydrophobicity surface on binding affinity of PCP-like compounds using machine learning methods
Yoldaş, Mine; Alpaslan, Ferda Nur; Büyükbingöl, Erdem; Department of Computer Engineering (2011)
This study aims to predict the binding affinity of the PCP-like compounds by means of molecular hydrophobicity. Molecular hydrophobicity is an important property which aff ects the binding affinity of molecules. The values of molecular hydrophobicity of molecules are obtained on three-dimensional coordinate system. Our aim is to reduce the number of points on the hydrophobicity surface of the molecules. This is modeled by using self organizing maps (SOM) and k-means clustering. The feature sets obtained fro...
Investigating the malleability of RNA aptamers
İlgü, Müslüm; Lamm, Monica H.; Nilsen-Hamilton, Marit (Elsevier BV, 2013-09-15)
Aptamers are short, single-stranded nucleic acids with structures that frequently change upon ligand binding and are sensitive to the ionic environment. To achieve facile application of aptamers in controlling cellular activities, a better understanding is needed of aptamer ligand binding parameters, structures, intramolecular mobilities and how these structures adapt to different ionic environments with consequent effects on their ligand binding characteristics. Here we discuss the integration of biochemic...
Citation Formats
D. B. E. ODTÜ, “Predictingprotein-protein interactions usingGCN-based encoder-decoder combined with a transformer,” Erdemli, Mersin, TÜRKİYE, 2022, p. 2119, Accessed: 00, 2023. [Online]. Available: