Article
Physics, Multidisciplinary
Hoyeon Jeong, Yoonbee Kim, Yi-Sue Jung, Dae Ryong Kang, Young-Rae Cho
Summary: Functional modules can be predicted using genome-wide protein-protein interactions based on graph clustering algorithms. Graph entropy (GE) is a novel metric for evaluating the quality of clusters in complex networks. The GE algorithm is more accurate in overlapping clusters compared to other competitive methods, confirming the biological feasibility of proteins within identified clusters and revealing new proteins for additional gene ontology annotations.
Article
Genetics & Heredity
Rongquan Wang, Huimin Ma, Caixia Wang
Summary: In this study, we developed a novel improved memetic algorithm (IMA) to detect protein complexes in protein-protein interaction networks. The IMA combines topological and biological properties and outperforms existing state-of-the-art techniques.
FRONTIERS IN GENETICS
(2021)
Article
Biochemical Research Methods
Hongwei Chen, Yunpeng Cai, Chaojie Ji, Gurudeeban Selvaraj, Dongqing Wei, Hongyan Wu
Summary: We propose an adaptive convolution graph network, AdaPPI, to predict protein functional modules in protein-protein interaction networks. By integrating protein gene ontology attributes and network topology, our framework outperforms existing methods in finding functional modules.
BRIEFINGS IN BIOINFORMATICS
(2023)
Article
Genetics & Heredity
Rongquan Wang, Huimin Ma, Caixia Wang
Summary: This study proposes an Ensemble Learning Framework (ELF-DPC) for detecting protein complexes within protein-protein interaction networks. The framework constructs a weighted PPI network by combining topological and biological information, mines protein complex cores using a designed strategy, and obtains an ensemble learning model by integrating structural modularity and a trained voting regressor model. Experimental results show that ELF-DPC outperforms state-of-the-art approaches and can detect biologically meaningful protein complexes.
FRONTIERS IN GENETICS
(2022)
Article
Biology
Kamal Berahmand, Elahe Nasiri, Rojiar Pir Mohammadiani, Yuefeng Li
Summary: The paper introduces a new spectral clustering method named TADWSC for identifying protein complexes in attributed networks. By combining topological structure and node features, the method improves the accuracy of protein complexes through calculating embedding vectors and the affinity matrix. The proposed method shows unexpectedly good performance compared to existing state-of-the-art methods in both real protein network datasets and synthetic networks.
COMPUTERS IN BIOLOGY AND MEDICINE
(2021)
Article
Biochemical Research Methods
Xiaoke Ma, Wei Zhao, Wenming Wu
Summary: Multi-layer networks provide an effective tool to model complex systems with multiple interactions. Graph clustering in multi-layer networks is challenging due to the difficulty in balancing cluster connectivity and layer connections. To address this, a novel algorithm called LSNMF is proposed to identify layer-specific modules in multi-layer networks. LSNMF first extracts vertex features using NMF and decomposes them into common and specific components, with orthogonality constraint imposed on the specific components. Extensive experiments show that LSNMF outperforms existing baselines and efficiently extracts stage-specific modules associated with known functions and survival time of patients.
IEEE-ACM TRANSACTIONS ON COMPUTATIONAL BIOLOGY AND BIOINFORMATICS
(2023)
Article
Biochemistry & Molecular Biology
Sara Omranian, Angela Angeleska, Zoran Nikoloski
Summary: GCC-v is an efficient, parameter-free algorithm that accurately predicts protein complexes, outperforming twelve state-of-the-art methods in multiple experimental scenarios. Its robustness to network perturbations is demonstrated in pan-plant PPI networks and Arabidopsis thaliana, highlighting its potential for impact assessment on predicted protein complexes.
COMPUTATIONAL AND STRUCTURAL BIOTECHNOLOGY JOURNAL
(2021)
Article
Biochemistry & Molecular Biology
Hang Zhou, Weikun Wang, Jiayun Jin, Zengwei Zheng, Binbin Zhou
Summary: This paper presents a comparative study of various graph neural networks for protein-protein interaction prediction. The experimental results show that hyperbolic graph neural networks tend to have a better performance than the other methods on the protein-related datasets.
Article
Biochemical Research Methods
Sridevi Maharaj, Taotao Qian, Zarin Ohiba, Wayne Hayes
Summary: The joint distribution of degree products and common neighbors has a greater impact on PPI edge connectivity than their individual distributions, leading to the introduction of two new models (CN and STICKY-CN). The inclusion of CN into STICKY-CN makes it the best overall fit for PPI networks as it is a good fit locally and globally.
IEEE-ACM TRANSACTIONS ON COMPUTATIONAL BIOLOGY AND BIOINFORMATICS
(2021)
Article
Biochemistry & Molecular Biology
Barnali Das, Pralay Mitra
Summary: The study develops ProMoCell and ProModb, which are used for clustering and storing cellular-level functional modules respectively, utilizing real-time web scraping and data from multiple online sources. These web services are synchronized with the KEGG pathway database and provide users with a comprehensive and efficient tool for browsing and extracting functional modules and interaction network data of cells. These services are of significant importance in pharmacological research.
JOURNAL OF MOLECULAR MODELING
(2022)
Article
Biochemical Research Methods
Yuanyuan Chen, Xiaodan Fan, Cong Pian
Summary: This article introduces a method for identifying functional (or disease-relevant) modules using gene expression data by integrating gene interaction networks and energy minimization with graph cuts method. The method is successful in identifying disease-relevant modules and performs well in real experiments.
CURRENT BIOINFORMATICS
(2021)
Article
Computer Science, Information Systems
Suyu Mei
Summary: In this study, a computational framework that combines supervised learning and dense subgraphs discovery is proposed to predict protein complexes. The framework reconstructs protein co-complex networks and infers complexes through effective clustering methods. Empirical studies show that the framework outperforms existing methods and provides biologically relevant protein complexes.
FRONTIERS OF COMPUTER SCIENCE
(2022)
Article
Biochemical Research Methods
Michela Quadrini, Sebastian Daberdaku, Carlo Ferrari
Summary: Protein-protein interactions play crucial roles in life processes. This study proposes a new abstraction method for protein structure and utilizes graph convolutional networks to predict protein interfaces. The results show that this method outperforms other competitors on structurally similar molecules.
BMC BIOINFORMATICS
(2022)
Article
Biochemistry & Molecular Biology
Jun-Xiao Ma, Yi Yang, Guang Li, Bin-Guang Ma
Summary: Symbiotic nitrogen fixation plays a crucial role in the nitrogen biogeochemical cycles and is the main nitrogen source for the biosphere. The molecular mechanisms of communication between rhizobia and host plants have been extensively studied, with a growing demand for integrated multiomics information. A computational framework was presented to study the protein-protein interaction network in the symbiosis system of B. diazoefficiens USDA110, revealing conserved functional modules and key protein hubs.
INTERNATIONAL JOURNAL OF MOLECULAR SCIENCES
(2021)
Article
Biology
Mehmet Burak Koca, Esmaeil Nourani, Ferda Abbasoglu, Ilknur Karadeniz, Fatih Erdogan Sevilgen
Summary: This study presents a three-stage machine learning pipeline for generating and using hybrid embeddings for PHI prediction. The method extracts numerical features from amino acid sequences and uses graph neural networks for training, ultimately utilizing hybrid protein embeddings for prediction.
COMPUTATIONAL BIOLOGY AND CHEMISTRY
(2022)