Search

Summary: Identifying natural or synthetic compounds in foods and assaying their bioactivities have significantly contributed to promoting human health. In this work, we propose a machine learning framework to predict 101 classes of health effects of food compounds at a large scale. In this framework, random undersampling boosting (RUSBoost) is used as base learners to tackle the problem of skewed class distributions and MACCSKeys similarity spectra are proposed as a feature engineering strategy to represent chemical molecules including food compounds, natural products and drugs. Computational results show that RUSBoost learners encouragingly reduce model biases, and that the knowledge learnt from food compounds is well transferable to natural products (0.8406–0.9040 recall rates for antibacterial, antivirals, pesticide and anticancer effects) and drugs (0.789–0.9690 recall rates for antibacterial, antiviral, antineoplastic and analgesic effects). and. Dozens of novel effects have been validated in recent literature. These pieces of evidence show that the proposed framework could help us to find lead compounds from food as potential pharmaceuticals and repurpose drugs for anticancer, antiviral or antibacterial therapies. Finally, we use the proposed framework to predict beneficial and risky health effects of food flavour compounds as case studies for recipe composing. [ABSTRACT FROM AUTHOR]

Abstract Background Prediction of protein localization in subnuclear organelles is more challenging than general protein subcelluar localization. There are only three computational models for protein subnuclear localization thus far, to the best of our knowledge. Two models were based on protein primary sequence only. The first model assumed homogeneous amino acid substitution pattern across all protein sequence residue sites and used BLOSUM62 to encode k-mer of protein sequence. Ensemble of SVM based on different k-mers drew the final conclusion, achieving 50% overall accuracy. The simplified assumption did not exploit protein sequence profile and ignored the fact of heterogeneous amino acid substitution patterns across sites. The second model derived the PsePSSM feature representation from protein sequence by simply averaging the profile PSSM and combined the PseAA feature representation to construct a kNN ensemble classifier Nuc-PLoc, achieving 67.4% overall accuracy. The two models based on protein primary sequence only both achieved relatively poor predictive performance. The third model required that GO annotations be available, thus restricting the model's applicability. Methods In this paper, we only use the amino acid information of protein sequence without any other information to design a widely-applicable model for protein subnuclear localization. We use K-spectrum kernel to exploit the contextual information around an amino acid and the conserved motif information. Besides expanding window size, we adopt various amino acid classification approaches to capture diverse aspects of amino acid physiochemical properties. Each amino acid classification generates a series of spectrum kernels based on different window size. Thus, (I) window expansion can capture more contextual information and cover size-varying motifs; (II) various amino acid classifications can exploit multi-aspect biological information from the protein sequence. Finally, we combine all the spectrum kernels by simple addition into one single kernel called SpectrumKernel+ for protein subnuclear localization. Results We conduct the performance evaluation experiments on two benchmark datasets: Lei and Nuc-PLoc. Experimental results show that SpectrumKernel+ achieves substantial performance improvement against the previous model Nuc-PLoc, with overall accuracy 83.47% against 67.4%; and 71.23% against 50% of Lei SVM Ensemble, against 66.50% of Lei GO SVM Ensemble. Conclusion The method SpectrumKernel+ can exploit rich amino acid information of protein sequence by embedding into implicit size-varying motifs the multi-aspect amino acid physiochemical properties captured by amino acid classification approaches. The kernels derived from diverse amino acid classification approaches and different sizes of k-mer are summed together for data integration. Experiments show that the method SpectrumKernel+ significantly outperforms the existing models for protein subnuclear localization.

Abstract Background Prediction of protein subcellular localization generally involves many complex factors, and using only one or two aspects of data information may not tell the true story. For this reason, some recent predictive models are deliberately designed to integrate multiple heterogeneous data sources for exploiting multi-aspect protein feature information. Gene ontology, hereinafter referred to as GO, uses a controlled vocabulary to depict biological molecules or gene products in terms of biological process, molecular function and cellular component. With the rapid expansion of annotated protein sequences, gene ontology has become a general protein feature that can be used to construct predictive models in computational biology. Existing models generally either concatenated the GO terms into a flat binary vector or applied majority-vote based ensemble learning for protein subcellular localization, both of which can not estimate the individual discriminative abilities of the three aspects of gene ontology. Results In this paper, we propose a Gene Ontology Based Transfer Learning Model (GO-TLM) for large-scale protein subcellular localization. The model transfers the signature-based homologous GO terms to the target proteins, and further constructs a reliable learning system to reduce the adverse affect of the potential false GO terms that are resulted from evolutionary divergence. We derive three GO kernels from the three aspects of gene ontology to measure the GO similarity of two proteins, and derive two other spectrum kernels to measure the similarity of two protein sequences. We use simple non-parametric cross validation to explicitly weigh the discriminative abilities of the five kernels, such that the time & space computational complexities are greatly reduced when compared to the complicated semi-definite programming and semi-indefinite linear programming. The five kernels are then linearly merged into one single kernel for protein subcellular localization. We evaluate GO-TLM performance against three baseline models: MultiLoc, MultiLoc-GO and Euk-mPLoc on the benchmark datasets the baseline models adopted. 5-fold cross validation experiments show that GO-TLM achieves substantial accuracy improvement against the baseline models: 80.38% against model Euk-mPLoc 67.40% with 12.98% substantial increase; 96.65% and 96.27% against model MultiLoc-GO 89.60% and 89.60%, with 7.05% and 6.67% accuracy increase on dataset MultiLoc plant and dataset MultiLoc animal, respectively; 97.14%, 95.90% and 96.85% against model MultiLoc-GO 83.70%, 90.10% and 85.70%, with accuracy increase 13.44%, 5.8% and 11.15% on dataset BaCelLoc plant, dataset BaCelLoc fungi and dataset BaCelLoc animal respectively. For BaCelLoc independent sets, GO-TLM achieves 81.25%, 80.45% and 79.46% on dataset BaCelLoc plant holdout, dataset BaCelLoc plant holdout and dataset BaCelLoc animal holdout, respectively, as compared against baseline model MultiLoc-GO 76%, 60.00% and 73.00%, with accuracy increase 5.25%, 20.45% and 6.46%, respectively. Conclusions Since direct homology-based GO term transfer may be prone to introducing noise and outliers to the target protein, we design an explicitly weighted kernel learning system (called Gene Ontology Based Transfer Learning Model, GO-TLM) to transfer to the target protein the known knowledge about related homologous proteins, which can reduce the risk of outliers and share knowledge between homologous proteins, and thus achieve better predictive performance for protein subcellular localization. Cross validation and independent test experimental results show that the homology-based GO term transfer and explicitly weighing the GO kernels substantially improve the prediction performance.

Background: Bacterial invasive infection and host immune response is fundamental to the understanding of pathogen pathogenesis and the discovery of effective therapeutic drugs. However, there are very few experimental studies on the signaling cross-talks between bacteria and human host to date. Methods: In this work, taking M. tuberculosis H37Rv (MTB) that is co-evolving with its human host as an example, we propose a general computational framework that exploits the known bacterial pathogen protein interaction networks in STRING database to predict pathogen-host protein interactions and their signaling cross-talks. In this framework, significant interlogs are derived from the known pathogen protein interaction networks to train a predictive l2-regularized logistic regression model. Results: The computational results show that the proposed method achieves excellent performance of cross validation as well as low predicted positive rates on the less significant interlogs and non-interlogs, indicating a low risk of false discovery. We further conduct gene ontology (GO) and pathway enrichment analyses of the predicted pathogen-host protein interaction networks, which potentially provides insights into the machinery that M. tuberculosis H37Rv targets human genes and signaling pathways. In addition, we analyse the pathogen-host protein interactions related to drug resistance, inhibition of which potentially provides an alternative solution to M. tuberculosis H37Rv drug resistance. Conclusions: The proposed machine learning framework has been verified effective for predicting bacteria-host protein interactions via known bacterial protein interaction networks. For a vast majority of bacterial pathogens that lacks experimental studies of bacteria-host protein interactions, this framework is supposed to achieve a general-purpose applicability. The predicted protein interaction networks between M. tuberculosis H37Rv and Homo sapiens, provided in the Additional files, promise to gain applications in the two fields: (1) providing an alternative solution to drug resistance; (2) revealing the patterns that M. tuberculosis H37Rv genes target human immune signaling pathways. [ABSTRACT FROM AUTHOR]

Understanding the physical arrangement of subunits within protein complexes potentially provides valuable clues about how the subunits work together and how the complexes function. The majority of recent research focuses on identifying protein complexes as a whole and seldom studies the inner structures within complexes. In this study, we propose a computational framework to predict direct contacts and substructures within protein complexes. In this framework, we first train a supervised learning model of l2-regularized logistic regression to learn the patterns of direct and indirect interactions within complexes, from where physical subunit interaction networks are predicted. Then, to infer substructures within complexes, we apply a graph clustering method (i.e., maximum modularity clustering (MMC)) and a gene ontology (GO) semantic similarity based functional clustering on partially- and fully-connected networks, respectively. Computational results show that the proposed framework achieves fairly good performance of cross validation and independent test in terms of detecting direct contacts between subunits. Functional analyses further demonstrate the rationality of partitioning the subunits into substructures via the MMC algorithm and functional clustering. [ABSTRACT FROM AUTHOR]

Protein-protein interaction (PPI) networks are naturally viewed as infrastructure to infer signalling pathways. The descriptors of signal events between two interacting proteins such as upstream/downstream signal flow, activation/inhibition relationship and protein modification are indispensable for inferring signalling pathways from PPI networks. However, such descriptors are not available in most cases as most PPI networks are seldom semantically annotated. In this work, we extend ℓ2-regularized logistic regression to the scenario of multi-label learning for predicting the activation/inhibition relationships in human PPI networks. The phenomenon that both activation and inhibition relationships exist between two interacting proteins is computationally modelled by multi-label learning framework. The problem of GO (gene ontology) sparsity is tackled by introducing the homolog knowledge as independent homolog instances. ℓ2-regularized logistic regression is accordingly adopted here to penalize the homolog noise and to reduce the computational complexity of the double-sized training data. Computational results show that the proposed method achieves satisfactory multi-label learning performance and outperforms the existing phenotype correlation method on the experimental data of Drosophila melanogaster. Several predictions have been validated against recent literature. The predicted activation/inhibition relationships in human PPI networks are provided in the supplementary file for further biomedical research. [ABSTRACT FROM AUTHOR]

Pathogen-host protein interactions are fundamental for pathogens to manipulate host signaling pathways and subvert host immune defense. For most pathogens, very few or no experimental studies have been conducted to investigate their signaling cross-talks with host. In this study, we propose a computational framework to validate the biological assumption that human protein-protein interaction (PPI) networks alone are sufficient to infer pathogen-host PPIs via pathogen functional mimicry. Pathogen functional mimicry assumes that a pathogen functionally mimics and substitutes host counterpart proteins in order for the pathogen to get involved in or hijack the host cellular processes. Through pathogen functional mimicry defined via gene ontology (GO) semantic similarity, we first use the known human PPIs as templates to infer pathogen-host PPIs, and the PPIs are further used as training data to build an l 2 -regularized logistic regression model for novel pathogen-host PPI prediction. Independent tests on the experimental data from human immunodeficiency virus and Francisella tularensis validate the effectiveness of the proposed pathogen functional mimicry technique. Performance comparisons also show that the proposed technique y excels the existing pathogen sequence mimicry approaches and transfer learning methods. The proposed framework provides a new avenue to study the experimentally less-studied pathogens in the worst scenarios that very few or no experimental pathogen-host PPIs are available. As two case studies, we apply the proposed framework to Salmonella typhimurium and Human respiratory syncytial virus to reconstruct the pathogen-host PPI networks and further investigate the interference of these two pathogens with human immune signaling and transcription regulatory system., Competing Interests: The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (© 2019 The Authors.)

Protein-protein interaction (PPI) networks are naturally viewed as infrastructure to infer signalling pathways. The descriptors of signal events between two interacting proteins such as upstream/downstream signal flow, activation/inhibition relationship and protein modification are indispensable for inferring signalling pathways from PPI networks. However, such descriptors are not available in most cases as most PPI networks are seldom semantically annotated. In this work, we extend ℓ 2 -regularized logistic regression to the scenario of multi-label learning for predicting the activation/inhibition relationships in human PPI networks. The phenomenon that both activation and inhibition relationships exist between two interacting proteins is computationally modelled by multi-label learning framework. The problem of GO (gene ontology) sparsity is tackled by introducing the homolog knowledge as independent homolog instances. ℓ 2 -regularized logistic regression is accordingly adopted here to penalize the homolog noise and to reduce the computational complexity of the double-sized training data. Computational results show that the proposed method achieves satisfactory multi-label learning performance and outperforms the existing phenotype correlation method on the experimental data of Drosophila melanogaster. Several predictions have been validated against recent literature. The predicted activation/inhibition relationships in human PPI networks are provided in the supplementary file for further biomedical research.