GlyNet: a multi-task neural network for predicting protein-glycan interactions.
Ontology highlight
ABSTRACT:
SUBMITTER: Carpenter EJ
PROVIDER: S-EPMC9172296 | biostudies-literature |
REPOSITORIES: biostudies-literature
Similar Datasets
Project description:Machine learning methods for protein function prediction are urgently needed, especially now that a substantial fraction of known sequences remains unannotated despite the extensive use of functional assignments based on sequence similarity. One major bottleneck supervised learning faces in protein function prediction is the structured, multi-label nature of the problem, because biological roles are represented by lists of terms from hierarchically organised controlled vocabularies such as the Gene Ontology. In this work, we build on recent developments in the area of deep learning and investigate the usefulness of multi-task deep neural networks (MTDNN), which consist of upstream shared layers upon which are stacked in parallel as many independent modules (additional hidden layers with their own output units) as the number of output GO terms (the tasks). MTDNN learns individual tasks partially using shared representations and partially from task-specific characteristics. When no close homologues with experimentally validated functions can be identified, MTDNN gives more accurate predictions than baseline methods based on annotation frequencies in public databases or homology transfers. More importantly, the results show that MTDNN binary classification accuracy is higher than alternative machine learning-based methods that do not exploit commonalities and differences among prediction tasks. Interestingly, compared with a single-task predictor, the performance improvement is not linearly correlated with the number of tasks in MTDNN, but medium size models provide more improvement in our case. One of advantages of MTDNN is that given a set of features, there is no requirement for MTDNN to have a bootstrap feature selection procedure as what traditional machine learning algorithms do. Overall, the results indicate that the proposed MTDNN algorithm improves the performance of protein function prediction. On the other hand, there is still large room for deep learning techniques to further enhance prediction ability.
| S-EPMC5995439 | biostudies-literature
Project description:Motivation:Triplet amino acids have successfully been included in feature selection to predict human-HPV protein-protein interactions (PPI). The utility of supervised learning methods is curtailed due to experimental data not being available in sufficient quantities. Improvements in machine learning techniques and features selection will enhance the study of PPI between host and pathogen. Results:We present a comparison of a neural network model versus SVM for prediction of host-pathogen PPI based on a combination of features including: amino acid quadruplets, pairwise sequence similarity, and human interactome properties. The neural network and SVM were implemented using Python Sklearn library. The neural network model using quadruplet features and other network features outperformance the SVM model. The models are tested against published predictors and then applied to the human-B.anthracis case. Gene ontology term enrichment analysis identifies immunology response and regulation as functions of interacting proteins. For prediction of Human-viral PPI, our model (neural network) is a significant improvement in overall performance compared to a predictor using the triplets feature and achieves a good accuracy in predicting human-B.anthracis PPI. Availability and implementation:All code can be downloaded from ftp://ftp.sanbi.ac.za/machine_learning/. Supplementary information:Supplementary data are available at Bioinformatics online.
| S-EPMC6289132 | biostudies-literature
Project description:Named Entity Recognition (NER) is a key task in biomedical text mining. Accurate NER systems require task-specific, manually-annotated datasets, which are expensive to develop and thus limited in size. Since such datasets contain related but different information, an interesting question is whether it might be possible to use them together to improve NER performance. To investigate this, we develop supervised, multi-task, convolutional neural network models and apply them to a large number of varied existing biomedical named entity datasets. Additionally, we investigated the effect of dataset size on performance in both single- and multi-task settings.We present a single-task model for NER, a Multi-output multi-task model and a Dependent multi-task model. We apply the three models to 15 biomedical datasets containing multiple named entities including Anatomy, Chemical, Disease, Gene/Protein and Species. Each dataset represent a task. The results from the single-task model and the multi-task models are then compared for evidence of benefits from Multi-task Learning. With the Multi-output multi-task model we observed an average F-score improvement of 0.8% when compared to the single-task model from an average baseline of 78.4%. Although there was a significant drop in performance on one dataset, performance improves significantly for five datasets by up to 6.3%. For the Dependent multi-task model we observed an average improvement of 0.4% when compared to the single-task model. There were no significant drops in performance on any dataset, and performance improves significantly for six datasets by up to 1.1%. The dataset size experiments found that as dataset size decreased, the multi-output model's performance increased compared to the single-task model's. Using 50, 25 and 10% of the training data resulted in an average drop of approximately 3.4, 8 and 16.7% respectively for the single-task model but approximately 0.2, 3.0 and 9.8% for the multi-task model.Our results show that, on average, the multi-task models produced better NER results than the single-task models trained on a single NER dataset. We also found that Multi-task Learning is beneficial for small datasets. Across the various settings the improvements are significant, demonstrating the benefit of Multi-task Learning for this task.
| S-EPMC5558737 | biostudies-other
Project description:Drug-target interactions (DTIs) prediction is an important step in drug discovery. As traditional biological experiments or high-throughput screening are high cost and time-consuming, many deep learning models have been developed. Overfitting must be avoided when training deep learning models. We propose a simple framework, called OverfitDTI, for DTI prediction. In OverfitDTI, a deep neural network (DNN) model is overfit to sufficiently learn the features of the chemical space of drugs and the biological space of targets. The weights of trained DNN model form an implicit representation of the nonlinear relationship between drugs and targets. Performance of OverfitDTI on three public datasets showed that the overfit DNN models fit the nonlinear relationship with high accuracy. We identified fifteen compounds that interacted with TEK, a receptor tyrosine kinase contributing to vascular homeostasis, and the predicted AT9283 and dorsomorphin were experimentally demonstrated as inhibitors of TEK in human umbilical vein endothelial cells (HUVECs).
| S-EPMC10480310 | biostudies-literature
Project description:BACKGROUND: Domains are the basic functional units of proteins. It is believed that protein-protein interactions are realized through domain interactions. Revealing multi-domain cooperation can provide deep insights into the essential mechanism of protein-protein interactions at the domain level and be further exploited to improve the accuracy of protein interaction prediction. RESULTS: In this paper, we aim to identify cooperative domains for protein interactions by extending two-domain interactions to multi-domain interactions. Based on the high-throughput experimental data from multiple organisms with different reliabilities, the interactions of domains were inferred by a Linear Programming algorithm with Multi-domain pairs (LPM) and an Association Probabilistic Method with Multi-domain pairs (APMM). Experimental results demonstrate that our approach not only can find cooperative domains effectively but also has a higher accuracy for predicting protein interaction than the existing methods. Cooperative domains, including strongly cooperative domains and superdomains, were detected from major interaction databases MIPS and DIP, and many of them were verified by physical interactions from the crystal structures of protein complexes in PDB which provide intuitive evidences for such cooperation. Comparison experiments in terms of protein/domain interaction prediction justified the benefit of considering multi-domain cooperation. CONCLUSION: From the computational viewpoint, this paper gives a general framework to predict protein interactions in a more accurate manner by considering the information of both multi-domains and multiple organisms, which can also be applied to identify cooperative domains, to reconstruct large complexes and further to annotate functions of domains. Supplementary information and software are provided in http://intelligent.eic.osaka-sandai.ac.jp/chenen/MDCinfer.htm and http://zhangroup.aporc.org/bioinfo/MDCinfer.
| S-EPMC2222654 | biostudies-literature
Project description:MotivationReconstruction of the network-level evolutionary history of protein-protein interactions provides a principled way to relate interactions in several present-day networks. Here, we present a general framework for inferring such histories and demonstrate how it can be used to determine what interactions existed in the ancestral networks, which present-day interactions we might expect to exist based on evolutionary evidence and what information extant networks contain about the order of ancestral protein duplications.ResultsOur framework characterizes the space of likely parsimonious network histories. It results in a structure that can be used to find probabilities for a number of events associated with the histories. The framework is based on a directed hypergraph formulation of dynamic programming that we extend to enumerate many optimal and near-optimal solutions. The algorithm is applied to reconstructing ancestral interactions among bZIP transcription factors, imputing missing present-day interactions among the bZIPs and among proteins from five herpes viruses, and determining relative protein duplication order in the bZIP family. Our approach more accurately reconstructs ancestral interactions than existing approaches. In cross-validation tests, we find that our approach ranks the majority of the left-out present-day interactions among the top 2 and 17% of possible edges for the bZIP and herpes networks, respectively, making it a competitive approach for edge imputation. It also estimates relative bZIP protein duplication orders, using only interaction data and phylogenetic tree topology, which are significantly correlated with sequence-based estimates.AvailabilityThe algorithm is implemented in C++, is open source and is available at http://www.cs.cmu.edu/ckingsf/software/parana2.Supplementary informationSupplementary data are available at Bioinformatics online.
| S-EPMC3694678 | biostudies-literature
Project description:Protein interactions play a vital part in the function of a cell. As experimental techniques for detection and validation of protein interactions are time consuming, there is a need for computational methods for this task. Protein interactions appear to form a network with a relatively high degree of local clustering. In this paper we exploit this clustering by suggesting a score based on triplets of observed protein interactions. The score utilises both protein characteristics and network properties. Our score based on triplets is shown to complement existing techniques for predicting protein interactions, outperforming them on data sets which display a high degree of clustering. The predicted interactions score highly against test measures for accuracy. Compared to a similar score derived from pairwise interactions only, the triplet score displays higher sensitivity and specificity. By looking at specific examples, we show how an experimental set of interactions can be enriched and validated. As part of this work we also examine the effect of different prior databases upon the accuracy of prediction and find that the interactions from the same kingdom give better results than from across kingdoms, suggesting that there may be fundamental differences between the networks. These results all emphasize that network structure is important and helps in the accurate prediction of protein interactions. The protein interaction data set and the program used in our analysis, and a list of predictions and validations, are available at http://www.stats.ox.ac.uk/bioinfo/resources/PredictingInteractions.
| S-EPMC2435280 | biostudies-literature
Project description:MotivationConvolutional neural networks (CNN) have outperformed conventional methods in modeling the sequence specificity of DNA-protein binding. Yet inappropriate CNN architectures can yield poorer performance than simpler models. Thus an in-depth understanding of how to match CNN architecture to a given task is needed to fully harness the power of CNNs for computational biology applications.ResultsWe present a systematic exploration of CNN architectures for predicting DNA sequence binding using a large compendium of transcription factor datasets. We identify the best-performing architectures by varying CNN width, depth and pooling designs. We find that adding convolutional kernels to a network is important for motif-based tasks. We show the benefits of CNNs in learning rich higher-order sequence features, such as secondary motifs and local sequence context, by comparing network performance on multiple modeling tasks ranging in difficulty. We also demonstrate how careful construction of sequence benchmark datasets, using approaches that control potentially confounding effects like positional or motif strength bias, is critical in making fair comparisons between competing methods. We explore how to establish the sufficiency of training data for these learning tasks, and we have created a flexible cloud-based framework that permits the rapid exploration of alternative neural network architectures for problems in computational biology.Availability and implementationAll the models analyzed are available at http://cnn.csail.mit.eduContactgifford@mit.eduSupplementary informationSupplementary data are available at Bioinformatics online.
| S-EPMC4908339 | biostudies-literature
Project description:The accurate identification of DNA sequence variants is an important, but challenging task in genomics. It is particularly difficult for single molecule sequencing, which has a per-nucleotide error rate of ~5-15%. Meeting this demand, we developed Clairvoyante, a multi-task five-layer convolutional neural network model for predicting variant type (SNP or indel), zygosity, alternative allele and indel length from aligned reads. For the well-characterized NA12878 human sample, Clairvoyante achieves 99.67, 95.78, 90.53% F1-score on 1KP common variants, and 98.65, 92.57, 87.26% F1-score for whole-genome analysis, using Illumina, PacBio, and Oxford Nanopore data, respectively. Training on a second human sample shows Clairvoyante is sample agnostic and finds variants in less than 2?h on a standard server. Furthermore, we present 3,135 variants that are missed using Illumina but supported independently by both PacBio and Oxford Nanopore reads. Clairvoyante is available open-source ( https://github.com/aquaskyline/Clairvoyante ), with modules to train, utilize and visualize the model.
| S-EPMC6397153 | biostudies-literature
Project description:Glycan microarrays provide a high-throughput means of profiling the interactions of glycan-binding proteins with their ligands. However, the construction of current glycan microarray platforms is time consuming and expensive. Here, we report a fast and cost-effective method for the assembly of cell-based glycan arrays to probe glycan-glycan-binding protein interactions directly on the cell surface. Chinese hamster ovary cell mutants with a narrow and relatively homogeneous repertoire of glycoforms serve as the foundation platforms to develop these arrays. Using recombinant glycosyltransferases, sialic acid, fucose, and analogs thereof are installed on cell-surface glycans to form cell-based arrays displaying diverse glycan epitopes that can be probed with glycan-binding proteins by flow cytometry. Using this platform, high-affinity glycan ligands are discovered for Siglec-15-a sialic acid-binding lectin involved in osteoclast differentiation. Incubating human osteoprogenitor cells with cells displaying a high-affinity Siglec-15 ligand impairs osteoclast differentiation, demonstrating the utility of this cell-based glycan array technology.
| S-EPMC5830402 | biostudies-literature