Protein Interaction Network Reconstruction Through Ensemble Deep Learning With Attention Mechanism

BackgroundIn a genetic interaction, the phenotype of a double mutant differs from the combined phenotypes of the underlying single mutants. When the single mutants have no growth defect, but the double mutant is lethal or exhibits slow growth, the interaction is termed synthetic lethality or synthetic fitness. These genetic interactions reveal gene redundancy and compensating pathways. Recently available large-scale data sets of genetic interactions and protein interactions in Saccharomyces cerevisiae provide a unique opportunity to elucidate the topological structure of biological pathways and how genes function in these pathways.ResultsWe have defined congruent genes as pairs of genes with similar sets of genetic interaction partners and constructed a genetic congruence network by linking congruent genes. By comparing path lengths in three types of networks (genetic interaction, genetic congruence, and protein interaction), we discovered that high genetic congruence not only exhibits correlation with direct protein interaction linkage but also exhibits commensurate distance with the protein interaction network. However, consistent distances were not observed between genetic and protein interaction networks. We also demonstrated that congruence and protein networks are enriched with motifs that indicate network transitivity, while the genetic network has both transitive (triangle) and intransitive (square) types of motifs. These results suggest that robustness of yeast cells to gene deletions is due in part to two complementary pathways (square motif) or three complementary pathways, any two of which are required for viability (triangle motif).ConclusionGenetic congruence is superior to genetic interaction in prediction of protein interactions and function associations. Genetically interacting pairs usually belong to parallel compensatory pathways, which can generate transitive motifs (any two of three pathways needed) or intransitive motifs (either of two pathways needed).

Protein–protein interactions (PPIs) are of great importance to understand genetic mechanisms, delineate disease pathogenesis, and guide drug design. With the increase of PPI data and development of machine learning technologies, prediction and identification of PPIs have become a research hotspot in proteomics. In this study, we propose a new prediction pipeline for PPIs based on gradient tree boosting (GTB). First, the initial feature vector is extracted by fusing pseudo amino acid composition (PseAAC), pseudo position-specific scoring matrix (PsePSSM), reduced sequence and index-vectors (RSIV), and autocorrelation descriptor (AD). Second, to remove redundancy and noise, we employ L1-regularized logistic regression (L1-RLR) to select an optimal feature subset. Finally, GTB-PPI model is constructed. Five-fold cross-validation showed that GTB-PPI achieved the accuracies of 95.15% and 90.47% on Saccharomyces cerevisiae and Helicobacter pylori datasets, respectively. In addition, GTB-PPI could be applied to predict the independent test datasets for Caenorhabditis elegans, Escherichia coli, Homo sapiens, and Mus musculus, the one-core PPI network for CD9, and the crossover PPI network for the Wnt-related signaling pathways. The results show that GTB-PPI can significantly improve accuracy of PPI prediction. The code and datasets of GTB-PPI can be downloaded from https://github.com/QUST-AIBBDRC/GTB-PPI/.

BackgroundWe have recently presented Schizophrenia Interactome, i.e., the network of protein-protein interactions (PPIs) of schizophrenia associated genes. PPIs predicted by our High-precision Protein-Protein Interaction Prediction model (HiPPIP) which uses machine learning to classify features of protein-pairs such as colocalization, coexpression, common molecular functions and biological processes could explain the apparent discordance between modern and historical genetic basis of Schizophrenia (published in npj Schizophrenia), and also were instrumental in discovering that OASL interacts DDX58 to activate the RIG-I immunity pathway during viral infection. These novel predicted PPIs were found to be highly accurate based on computational and experimental validations, and gave insights into possible functions of SZ genes that previously had no-known functional information. Even a single novel PPI can have enormous impact on advancement of biology, when translated effectively. How can we ensure that 500+ of these novel PPIs of schizophrenia interactome are translated effectively?MethodsWe developed a platform for Massive Open Online Discovery for Schizophrenia Research, or “MOOD for Schizophrenia Research”, that brings together trained biologists and bioinformaticians including those who are currently not affiliated with research labs (non-research students, PhDs who gave up science careers for administrative/corporate jobs or to take care of families, etc), as well as scientists who are active researchers, to hypothesize, discuss and prioritize the novel PPIs. Hypotheses are written as nanopublications with authorship credit. We are developing a number of features on the portal that allow and encourage scientists to create knowledge around the predicted PPIs and be recognized and given credit.ResultsThe first version of our website that disseminates the Schizophrenia Interactome is receiving hundreds of unique users each month. We have since developed MOOD for Schizophrenia Research and will present the key features of the website in this work. Each PPI can be viewed, researched on and written about, by participating scientists. Comprehensive information about the proteins in the PPI regarding their known functions, pathways, diseases and drug associations, is made readily available to scientists, allowing them to hypothesize the importance of the specific PPI. We present methods that we employ to promote collaboration in this work. Initially, the portal starts with a few members and it grows through referral webs (i.e. current members invite new members). The portal has a number of features that recognize and thus entice users to participate. We will also present the user feedback and participation.DiscussionPPIs are central to cellular systems. Yet less than 10% of estimated PPIs are known today. Thus, the computationally predicted PPIs which are deemed accurate, can accelerate advancement of schizophrenia biology research. The time is ripe to benefit from computer science and information technologies methods for not only discovering aspects of computational biology but to create new mechanisms to promote online collaboration to achieve big things as a summation of nanocontributions. The knowledge potential generated through this system would aid various principal investigators in well established (but ill funded) research labs by giving them access to bioinformaticians and biologists around the world who are eager be recognized for their contributions. We present here, not merely a website but a novel approach to promote collaborative research between people with heterogeneous skills and commitments by benefiting from the untapped talent of researchers around the world.

Radiologists often face challenges in differentiating benign from malignant sacral bone lesions due to their similar imaging characteristics. This study aimed to develop an ensemble deep learning (DL) model that can preoperatively distinguish between benign and malignant sacral tumors using noncontrast computed tomography images. Preoperative sacral CT scans from 569 patients with confirmed sacral lesions were analyzed. Data from Center 1 were utilized in model development and internal test via fivefold cross-validation, and those from Centers 2 and 3 were employed in external test. Various ensemble models combining human-readable interpretation and DL were developed. The diagnostic performance of the models and radiologists was assessed using metrics such as precision, recall, accuracy, area under the curve (AUC), F1 score, and confusion matrix. Furthermore, the clinical benefits derived from radiologists' interpretations and supported by the DL model were evaluated. The ensemble model, which integrates 3D-DenseNet121 with human interpretation, exhibited the most robust performance. The ensemble model demonstrated high performance on the internal and external test sets and achieved AUCs of 0.9139 and 0.8713, F1 scores of 0.9054 and 0.8571, precision of 0.9041 and 0.8824, recall of 0.9136 and 0.8333, and accuracy of 0.8630 and 0.8182, respectively. Across the external test cohort, all radiologists experienced improvements in AUC, accuracy, sensitivity, and specificity. Notably, junior radiologists demonstrated significant improvements compared with senior radiologists. The potential clinical application of the DL model lies in its capacity to considerably enhance the diagnostic efficiency of radiologists. This study presents the first ensemble deep learning model integrating 3D-DenseNet121 with radiologists' interpretation for preoperative differentiation of sacral tumors on noncontrast CT that improved diagnostic performance across all experience levels, particularly for junior radiologists. First artificial intelligence-radiologist ensemble for noncontrast computed tomography (NCCT)-based sacral tumor classification. Boosts all radiologists' performance, with the greatest gains for juniors, potentially reducing referrals. Enables reliable NCCT diagnosis, overcoming contrast/magnetic resonance imaging dependency in musculoskeletal oncology.

Pathogenic microorganisms exploit host cellular mechanisms and evade host defense mechanisms through molecular pathogen-host interactions (PHIs). Therefore, comprehensive analysis of these PHI networks should be an initial step for developing effective therapeutics against infectious diseases. Computational prediction of PHI data is gaining increasing demand because of scarcity of experimental data. Prediction of protein-protein interactions (PPIs) within PHI systems can be formulated as a classification problem, which requires the knowledge of non-interacting protein pairs. This is a restricting requirement since we lack datasets that report non-interacting protein pairs. In this study, we formulated the "computational prediction of PHI data" problem using kernel embedding of heterogeneous data. This eliminates the abovementioned requirement and enables us to predict new interactions without randomly labeling protein pairs as non-interacting. Domain-domain associations are used to filter the predicted results leading to 175 novel PHIs between 170 human proteins and 105 viral proteins. To compare our results with the state-of-the-art studies that use a binary classification formulation, we modified our settings to consider the same formulation. Detailed evaluations are conducted and our results provide more than 10 percent improvements for accuracy and AUC (area under the receiving operating curve) results in comparison with state-of-the-art methods.

Prediction of protein-protein interaction (PPI) is one of the most challenging problems in biology. Although great progress has been devoted to the development of methodology for predicting PPIs and PIN using machine learning methods, the problem is still far from being solved since the application of most existing methods is limited. In this study, we propose a method for PPI prediction based on amino acids differences between pairs of protein sequences. 10-fold cross-validation tests based on human PPI datasets with balanced positive-to-negative ratios indicate that it performs comparably well. Therefore, our finding suggests that amino acids differences of interacting protein pairs are relevant to the prediction of PPIs and hence provide important information on sequence-based encoding schemes.

Objectives: To design a new framework to efficiently parallelize the steps of VLASPD algorithm using a hybridized apriori and fp-growth on GPU; to implement the existing and proposed framework in CUDA;to improve the performance factors like computational time, memory and CPU utilization.Methods/Statistical Analysis: This paper proposes the acceleration of Protein-Protein Interactions (PPIs) prediction on Graphics Processing Units(GPUs). A GPU can provide more processing cores and computational power in the same cost as a CPU.Findings: The frequently occurring patterns in the protein sequences can be used for PPIs prediction.The moving of the approaches from fixed length to variable length lead to computational complexity but also is found to be advantageous.Applications/Improvements:Sequence biology is since being researched by various computer engineers, the GPUs can be employed for predicting various sequence interactions like DNA-Proteins, etc. Since the GPU runs the parallel code efficiently, the methodology can be further improved if efficiently parallelized.

Accurate finding of Knee Osteoarthritis (KOA) from structural Magnetic Resonance Imaging (MRI) is a difficult task and is greatly subject to user variation. Furthermore, the identification of knee osteoarthritis (KOA) from MRI scans presents a challenge due to the limited information available. A novel methodology using an ensemble Deep Learning algorithm, combining EfficientNet-B3 and ResNext-101 architectures, aims to forecast KOA advancement, bridging the identified gap in clinical trials. The study aims to develop a precise predictive model for knee osteoarthritis using advanced deep-learning architectures and structural MRI scan data. By utilizing an ensemble technique, the model's accuracy in predicting disease development is enhanced, surpassing the limitations of traditional biomarkers. The study used the Osteoarthritis Initiative dataset to develop an ensemble Deep Learning model that combined EfficientNet-B3 and ResNext-101 architectures. Techniques like cropping, gamma correction, and in-slice rotation were used to expand the dataset and improve the model's generalization capacity. The Deep Learning model demonstrated 93% validation accuracy on the OAI dataset, accurately capturing subtle patterns of knee osteoarthritis progression. Augmentation approaches enhanced its resilience. Our ensemble Deep Learning approach, using ResNext-101 and EfficientNet-B3 architectures, accurately predicts knee osteoarthritis courses using structural MRI data, demonstrating the importance of data augmentation for improved predictive tools.

Cliques (maximal complete subnets) in protein-protein interaction (PPI) network are an important resource used to analyze protein complexes and functional modules. Clique-based methods of predicting PPI complement the data defection from biological experiments. However, clique-based predicting methods only depend on the topology of network. The false-positive and false-negative interactions in a network usually interfere with prediction. Therefore, we propose a method combining clique-based method of prediction and gene ontology (GO) annotations to overcome the shortcoming and improve the accuracy of predictions. According to different GO correcting rules, we generate two predicted interaction sets which guarantee the quality and quantity of predicted protein interactions. The proposed method is applied to the PPI network from the Database of Interacting Proteins (DIP) and most of the predicted interactions are verified by another biological database, BioGRID. The predicted protein interactions are appended to the original protein network, which leads to clique extension and shows the significance of biological meaning.

Improved prediction of protein-protein interactions using novel negative samples, features, and an ensemble classifier.

Protein-protein interaction (PPI) prediction is a central task in achieving a better understanding of cellular and intracellular processes. Because high-throughput experimental methods are both expensive and time-consuming, and are also known of suffering from the problems of incompleteness and noise, many computational methods have been developed, with varied degrees of success. However, the inference of PPI network from multiple heterogeneous data sources remains a great challenge. In this work, we developed a novel method based on approximate Bayesian computation and modified differential evolution sampling (ABC-DEP) and regularized laplacian (RL) kernel. The method enables inference of PPI networks from topological properties and multiple heterogeneous features including gene expression and Pfam domain profiles, in forms of weighted kernels. The optimal weights are obtained by ABC-DEP, and the kernel fusion built based on optimal weights serves as input to RL to infer missing or new edges in the PPI network. Detailed comparisons with control methods have been made, and the results show that the accuracy of PPI prediction measured by AUC is increased by up to 23 %, as compared to a baseline without using optimal weights. The method can provide insights into the relations between PPIs and various feature kernels and demonstrates strong capability of predicting faraway interactions that cannot be well detected by traditional RL method.

Protein-Protein Interactions (PPIs) are fundamental means of functions and signalings in biological systems. The massive growth in demand and cost associated with experimental PPI studies calls for computational tools for automated prediction and understanding of PPIs. Despite recent progress, in silico methods remain inadequate in modeling the natural PPI hierarchy. Here we present a double-viewed hierarchical graph learning model, HIGH-PPI, to predict PPIs and extrapolate the molecular details involved. In this model, we create a hierarchical graph, in which a node in the PPI network (top outside-of-protein view) is a protein graph (bottom inside-of-protein view). In the bottom view, a group of chemically relevant descriptors, instead of the protein sequences, are used to better capture the structure-function relationship of the protein. HIGH-PPI examines both outside-of-protein and inside-of-protein of the human interactome to establish a robust machine understanding of PPIs. This model demonstrates high accuracy and robustness in predicting PPIs. Moreover, HIGH-PPI can interpret the modes of action of PPIs by identifying important binding and catalytic sites precisely. Overall, “HIGH-PPI [https://github.com/zqgao22/HIGH-PPI]” is a domain-knowledge-driven and interpretable framework for PPI prediction studies.

Many essential cellular functions are carried out by multi-protein complexes that can be characterized by their protein–protein interactions. The interactions between protein subunits are critically dependent on the strengths of their interactions and their cellular abundances, both of which span orders of magnitude. Despite many efforts devoted to the global discovery of protein complexes by integrating large-scale protein abundance and interaction features, there is still room for improvement. Here, we integrated >7000 quantitative proteomic samples with three published affinity purification/co-fractionation mass spectrometry datasets into a deep learning framework to predict protein–protein interactions (PPIs), followed by the identification of protein complexes using a two-stage clustering strategy. Our deep-learning-technique-based classifier significantly outperformed recently published machine learning prediction models and in the process captured 5010 complexes containing over 9000 unique proteins. The vast majority of proteins in our predicted complexes exhibited low or no tissue specificity, which is an indication that the observed complexes tend to be ubiquitously expressed throughout all cell types and tissues. Interestingly, our combined approach increased the model sensitivity for low abundant proteins, which amongst other things allowed us to detect the interaction of MCM10, which connects to the replicative helicase complex via the MCM6 protein. The integration of protein abundances and their interaction features using a deep learning approach provided a comprehensive map of protein–protein interactions and a unique perspective on possible novel protein complexes.

BackgroundThe bHLH (basic helix-loop-helix) transcription factor is one of the largest families of transcription factors in plants, containing a large number of members with diverse functions. Chinese jujube (Ziziphus jujuba Mill.) is the species with the highest economic value in the family Rhamnaceae. However, the characteristics of the bHLH family in the jujube genome are still unclear. Hence, ZjbHLHs were first searched at a genome-wide level, their expression levels under various conditions were investigated systematically, and their protein-protein interaction networks were predicted.ResultsWe identified 92 ZjbHLHs in the jujube genome, and these genes were classified into 16 classes according to bHLH domains. Ten ZjbHLHs with atypical bHLH domains were found. Seventy ZjbHLHs were mapped to but not evenly distributed on 12 pseudo- chromosomes. The domain sequences among ZjbHLHs were highly conserved, and their conserved residues were also identified. The tissue-specific expression of 37 ZjbHLH genes in jujube and wild jujube showed diverse patterns, revealing that these genes likely perform multiple functions. Many ZjbHLH genes were screened and found to be involved in flower and fruit development, especially in earlier developmental stages. A few genes responsive to phytoplasma invasion were also verified. Based on protein-protein interaction prediction and homology comparison, protein-protein interaction networks composed of 92 ZjbHLHs were also established.ConclusionsThis study provides a comprehensive bioinformatics analysis of 92 identified ZjbHLH genes. We explored their expression patterns in various tissues, the flowering process, and fruit ripening and under phytoplasma stress. The protein-protein interaction networks of ZjbHLHs provide valuable clues toward further studies of their biological functions.

Nowadays, multiple sources of information about proteins are available such as protein sequences, 3D structures, Gene Ontology (GO), etc. Most of the works on protein-protein interaction (PPI) identification had utilized these information about proteins, mainly sequence-based, but individually. The new advances in deep learning techniques allow us to leverage multiple sources/modalities of proteins, which complement each other. Some recent works have shown that multi-modal PPI models perform better than uni-modal approaches. This paper aims to investigate whether the performance of multi-modal PPI models is always consistent or depends on other factors such as dataset distribution, algorithms used to learn features, etc. We have used three modalities for this study: Protein sequence, 3D structure, and GO. Various techniques, including deep learning algorithms, are employed to extract features from multiple sources of proteins. These feature vectors from different modalities are then integrated in several combinations (bi-modal and tri-modal) to predict PPI. To conduct this study, we have used Human and S. cerevisiae PPI datasets. The obtained results demonstrate the potentiality of a multi-modal approach and deep learning techniques in predicting protein interactions. However, the predictive capability of a model for PPI depends on feature extraction methods as well. Also, increasing the modality does not always ensure performance improvement. In this study, the PPI model integrating two modalities outperforms the designed uni-modal and tri-modal PPI models.