On protocols and measures for the validation of supervised methods for the inference of biological networks

Heterogeneous network approaches to protein pathway prediction

Understanding protein-protein interactions (PPIs) and the pathways they comprise is essential for comprehending cellular functions and their links to specific phenotypes. Despite the prevalence of molecular data generated by high-throughput sequencing technologies, a significant gap remains in translating this data into functional information regarding the series of interactions that underlie phenotypic differences. In this review, we present an in-depth analysis of heterogeneous network methodologies for modeling protein pathways, highlighting the critical role of integrating multifaceted biological data. It outlines the process of constructing these networks, from data representation to machine learning-driven predictions and evaluations. The work underscores the potential of heterogeneous networks in capturing the complexity of proteomic interactions, thereby offering enhanced accuracy in pathway prediction. This approach not only deepens our understanding of cellular processes but also opens up new possibilities in disease treatment and drug discovery by leveraging the predictive power of comprehensive proteomic data analysis.

Risk factors for readmission after sepsis and its association with mortality

BACKGROUND

Sepsis is associated with an approximately 20 % 30-day readmission rate and with subsequent mortality.

OBJECTIVES

To determine the demographics, comorbidities that had been documented prior to sepsis onset, processes of care, commonly administered laboratory tests measured near discharge, and post-sepsis infections that may be associated with readmission and, secondarily, whether readmission is an independent risk factor for 90-day mortality.

METHODS

Using a database of patients who met Sepsis-3 criteria divided into Construction and Validation groups, we used logistic regression to estimate the factors independently associated with readmission within 30 days after discharge and proportional hazard regression to estimate the factors independently associated with 90-day mortality.

RESULTS

Of the 30,798 patients ≥ 18 years at our combined referral and community hospital and were discharged alive who met Sepsis-3 criteria between July 10, 2009 and September 7, 2019, 5943 (19 %) were readmitted within 30 days. Thirteen thousand, four hundred forty-four (44 %) of the patients were female, 25,293 (82 %) White, 3523 (11 %) Black, and the mean age was 59 ± 17 years. Among the readmitted patients, 894 (15 %) died within 90 days from the original discharge compared to 11 % (p < 0.001) who had not been readmitted. Seven comorbidities, five processes of care (presepsis platelet transfusion, postsepsis platelet transfusion, operation, ICU length of stay, and hospital length of stay), five culture results, two discharge laboratory values, and discharge location were associated with readmission. The model had good discrimination, 0.770 ± 0.004 (Construction Group) and 0.748 ± 0.006 (Validation Group) and good relevancy (area under the precision recall curve), 0.390 ± 0.004 (Construction group) and 0.476 ± 0.005 (Validation group). Readmission within 30 days was independently associated with a 56 % higher risk of death (HR=1.562, 95 % CI=1.434, 1.703, p < 0.001) within 90 days from discharge.

CONCLUSIONS

Comorbidities, abnormal laboratory values, processes of care, and post-sepsis onset culture results, but not demographic characteristics, were associated with 30-day readmission. Readmission was associated with 90-day mortality.

Nitrogen sensing and regulatory networks: it's about time and space

A plant's response to external and internal nitrogen signals/status relies on sensing and signaling mechanisms that operate across spatial and temporal dimensions. From a comprehensive systems biology perspective, this involves integrating nitrogen responses in different cell types and over long distances to ensure organ coordination in real time and yield practical applications. In this prospective review, we focus on novel aspects of nitrogen (N) sensing/signaling uncovered using temporal and spatial systems biology approaches, largely in the model Arabidopsis. The temporal aspects span: transcriptional responses to N-dose mediated by Michaelis-Menten kinetics, the role of the master NLP7 transcription factor as a nitrate sensor, its nitrate-dependent TF nuclear retention, its "hit-and-run" mode of target gene regulation, and temporal transcriptional cascade identified by "network walking." Spatial aspects of N-sensing/signaling have been uncovered in cell type-specific studies in roots and in root-to-shoot communication. We explore new approaches using single-cell sequencing data, trajectory inference, and pseudotime analysis as well as machine learning and artificial intelligence approaches. Finally, unveiling the mechanisms underlying the spatial dynamics of nitrogen sensing/signaling networks across species from model to crop could pave the way for translational studies to improve nitrogen-use efficiency in crops. Such outcomes could potentially reduce the detrimental effects of excessive fertilizer usage on groundwater pollution and greenhouse gas emissions.

Integrative approach for predicting drug-target interactions via matrix factorization and broad learning systems

In the drug discovery process, time and costs are the most typical problems resulting from the experimental screening of drug-target interactions (DTIs). To address these limitations, many computational methods have been developed to achieve more accurate predictions. However, identifying DTIs mostly rely on separate learning tasks with drug and target features that neglect interaction representation between drugs and target. In addition, the lack of these relationships may lead to a greatly impaired performance on the prediction of DTIs. Aiming at capturing comprehensive drug-target representations and simplifying the network structure, we propose an integrative approach with a convolution broad learning system for the DTI prediction (ConvBLS-DTI) to reduce the impact of the data sparsity and incompleteness. First, given the lack of known interactions for the drug and target, the weighted K-nearest known neighbors (WKNKN) method was used as a preprocessing strategy for unknown drug-target pairs. Second, a neighborhood regularized logistic matrix factorization (NRLMF) was applied to extract features of updated drug-target interaction information, which focused more on the known interaction pair parties. Then, a broad learning network incorporating a convolutional neural network was established to predict DTIs, which can make classification more effective using a different perspective. Finally, based on the four benchmark datasets in three scenarios, the ConvBLS-DTI's overall performance out-performed some mainstream methods. The test results demonstrate that our model achieves improved prediction effect on the area under the receiver operating characteristic curve and the precision-recall curve.

Improving Mortality Risk Prediction with Routine Clinical Data: A Practical Machine Learning Model Based on eICU Patients

Purpose

Mortality risk prediction helps clinicians make better decisions in patient healthcare. However, existing severity scoring systems or algorithms used in intensive care units (ICUs) often rely on laborious manual collection of complex variables and lack sufficient validation in diverse clinical environments, thus limiting their practical applicability. This study aims to evaluate the performance of machine learning models that utilize routinely collected clinical data for short-term mortality risk prediction.

Patients and Methods

Using the eICU Collaborative Research Database, we identified a cohort of 12,393 ICU patients, who were randomly divided into a training group and a validation group at a ratio of 9:1. The models utilized routine variables obtained from regular medical workflows, including age, gender, physiological measurements, and usage of vasoactive medications within a 24-hour period prior to patient discharge. Four different machine learning algorithms, namely logistic regression, random forest, extreme gradient boosting (XGboost), and artificial neural network were employed to develop the mortality risk prediction model. We compared the discrimination and calibration performance of these models in assessing mortality risk within 1-week time window.

Results

Among the tested models, the XGBoost algorithm demonstrated the highest performance, with an area under the receiver operating characteristic curve (AUROC) of 0.9702, an area under precision and recall curves (AUPRC) of 0.8517, and a favorable Brier score of 0.0259 for 24-hour mortality risk prediction. Although the model's performance decreased when considering larger time windows, it still achieved a comparable AUROC of 0.9184 and AUPRC of 0.5519 for 3-day mortality risk prediction.

Conclusion

The findings demonstrate the feasibility of developing a highly accurate and well-calibrated model based on the XGBoost algorithm for short-term mortality risk prediction with easily accessible and interpretative data. These results enhance confidence in the application of the machine learning model to clinical practice.

Explainable artificial intelligence-based prediction of poor neurological outcome from head computed tomography in the immediate post-resuscitation phase

Predicting poor neurological outcomes after resuscitation is important for planning treatment strategies. We constructed an explainable artificial intelligence-based prognostic model using head computed tomography (CT) scans taken immediately within 3 h of resuscitation from cardiac arrest and compared its predictive accuracy with that of previous methods using gray-to-white matter ratio (GWR). We included 321 consecutive patients admitted to our institution after resuscitation for out-of-hospital cardiopulmonary arrest with circulation resumption over 6 years. A machine learning model using head CT images with transfer learning was used to predict the neurological outcomes at 1 month. These predictions were compared with the predictions of GWR for multiple regions of interest in head CT using receiver operating characteristic (ROC)-area under curve (AUC) and precision recall (PR)-AUC. The regions of focus were visualized using a heatmap. Both methods had similar ROC-AUCs, but the machine learning model had a higher PR-AUC (0.73 vs. 0.58). The machine learning-focused area of interest for classification was the boundary between gray and white matter, which overlapped with the area of focus when diagnosing hypoxic- ischemic brain injury. The machine learning model for predicting poor outcomes had superior accuracy to conventional methods and could help optimize treatment.

PiLSL: pairwise interaction learning-based graph neural network for synthetic lethality prediction in human cancers

MOTIVATION

Synthetic lethality (SL) is a type of genetic interaction in which the simultaneous inactivation of two genes leads to cell death, while the inactivation of a single gene does not affect the cell viability. It can effectively expand the range of anti-cancer therapeutic targets. SL interactions are identified mainly by experimental screening and computational prediction. Recent machine-learning methods mostly learn the representation of each gene individually, ignoring the representation of the pairwise interaction between two genes. In addition, the mechanisms of SL, the key to translating SL into cancer therapeutics, are often unclear.

RESULTS

To fill the gaps, we propose a pairwise interaction learning-based graph neural network (GNN) named PiLSL to learn the representation of pairwise interaction between two genes for SL prediction. First, we construct an enclosing graph for each pair of genes from a knowledge graph. Secondly, we design an attentive embedding propagation layer in a GNN to discriminate the importance among the edges in the enclosing graph and to learn the latent features of the pairwise interaction from the weighted enclosing graph. Finally, we further fuse the latent features with explicit features extracted from multi-omics data to obtain powerful gene representations for SL prediction. Extensive experimental results demonstrate that PiLSL outperforms the best baseline by a large margin and generalizes well under three realistic scenarios. Besides, PiLSL provides an explanation of SL mechanisms via the weighted paths in the enclosing graphs by attention mechanism.

AVAILABILITY AND IMPLEMENTATION

Our source code is available at https://github.com/JieZheng-ShanghaiTech/PiLSL.

CIPHER-SC: Disease-Gene Association Inference Using Graph Convolution on a Context-Aware Network With Single-Cell Data

Inference of disease-gene associations helps unravel the pathogenesis of diseases and contributes to the treatment. Although many machine learning-based methods have been developed to predict causative genes, accurate association inference remains challenging. One major reason is the inaccurate feature selection and accumulation of error brought by commonly used multi-stage training architecture. In addition, the existing methods do not incorporate cell-type-specific information, thus fail to study gene functions at a higher resolution. Therefore, we introduce single-cell transcriptome data and construct a context-aware network to unbiasedly integrate all data sources. Then we develop a graph convolution-based approach named CIPHER-SC to realize a complete end-to-end learning architecture. Our approach outperforms four state-of-the-art approaches in five-fold cross-validations on three distinct test sets with the best AUC of 0.9501, demonstrating its stable ability either to predict the novel genes or to predict with genetic basis. The ablation study shows that our complete end-to-end design and unbiased data integration boost the performance from 0.8727 to 0.9443 in AUC. The addition of single-cell data further improves the prediction accuracy and makes our results be enriched for cell-type-specific genes. These results confirm the ability of CIPHER-SC to discover reliable disease genes. Our implementation is available at http://github.com/YidingZhang117/CIPHER-SC.

Network inference with Granger causality ensembles on single-cell transcriptomics

Cellular gene expression changes throughout a dynamic biological process, such as differentiation. Pseudotimes estimate cells' progress along a dynamic process based on their individual gene expression states. Ordering the expression data by pseudotime provides information about the underlying regulator-gene interactions. Because the pseudotime distribution is not uniform, many standard mathematical methods are inapplicable for analyzing the ordered gene expression states. Here we present single-cell inference of networks using Granger ensembles (SINGE), an algorithm for gene regulatory network inference from ordered single-cell gene expression data. SINGE uses kernel-based Granger causality regression to smooth irregular pseudotimes and missing expression values. It aggregates predictions from an ensemble of regression analyses to compile a ranked list of candidate interactions between transcriptional regulators and target genes. In two mouse embryonic stem cell differentiation datasets, SINGE outperforms other contemporary algorithms. However, a more detailed examination reveals caveats about poor performance for individual regulators and uninformative pseudotimes.

Machine Learning Predicts Drug Metabolism and Bioaccumulation by Intestinal Microbiota

Over 150 drugs are currently recognised as being susceptible to metabolism or bioaccumulation (together described as depletion) by gastrointestinal microorganisms; however, the true number is likely higher. Microbial drug depletion is often variable between and within individuals, depending on their unique composition of gut microbiota. Such variability can lead to significant differences in pharmacokinetics, which may be associated with dosing difficulties and lack of medication response. In this study, literature mining and unsupervised learning were used to curate a dataset of 455 drug-microbiota interactions. From this, 11 supervised learning models were developed that could predict drugs' susceptibility to depletion by gut microbiota. The best model, a tuned extremely randomised trees classifier, achieved performance metrics of AUROC: 75.1% ± 6.8; weighted recall: 79.2% ± 3.9; balanced accuracy: 69.0% ± 4.6; and weighted precision: 80.2% ± 3.7 when validated on 91 drugs. This machine learning model is the first of its kind and provides a rapid, reliable, and resource-friendly tool for researchers and industry professionals to screen drugs for susceptibility to depletion by gut microbiota. The recognition of drug-microbiome interactions can support successful drug development and promote better formulations and dosage regimens for patients.

Challenges and opportunities in network-based solutions for biological questions

Network biology is useful for modeling complex biological phenomena; it has attracted attention with the advent of novel graph-based machine learning methods. However, biological applications of network methods often suffer from inadequate follow-up. In this perspective, we discuss obstacles for contemporary network approaches—particularly focusing on challenges representing biological concepts, applying machine learning methods, and interpreting and validating computational findings about biology—in an effort to catalyze actionable biological discovery.

FARCI: Fast and Robust Connectome Inference

The inference of neuronal connectome from large-scale neuronal activity recordings, such as two-photon Calcium imaging, represents an active area of research in computational neuroscience. In this work, we developed FARCI (Fast and Robust Connectome Inference), a MATLAB package for neuronal connectome inference from high-dimensional two-photon Calcium fluorescence data. We employed partial correlations as a measure of the functional association strength between pairs of neurons to reconstruct a neuronal connectome. We demonstrated using in silico datasets from the Neural Connectomics Challenge (NCC) and those generated using the state-of-the-art simulator of Neural Anatomy and Optimal Microscopy (NAOMi) that FARCI provides an accurate connectome and its performance is robust to network sizes, missing neurons, and noise levels. Moreover, FARCI is computationally efficient and highly scalable to large networks. In comparison with the best performing connectome inference algorithm in the NCC, Generalized Transfer Entropy (GTE), and Fluorescence Single Neuron and Network Analysis Package (FluoroSNNAP), FARCI produces more accurate networks over different network sizes, while providing significantly better computational speed and scaling.

Machine Learning Uncovers Adverse Drug Effects on Intestinal Bacteria

The human gut microbiome, composed of trillions of microorganisms, plays an essential role in human health. Many factors shape gut microbiome composition over the life span, including changes to diet, lifestyle, and medication use. Though not routinely tested during drug development, drugs can exert profound effects on the gut microbiome, potentially altering its functions and promoting disease. This study develops a machine learning (ML) model to predict whether drugs will impair the growth of 40 gut bacterial strains. Trained on over 18,600 drug-bacteria interactions, 13 distinct ML models are built and compared, including tree-based, ensemble, and artificial neural network techniques. Following hyperparameter tuning and multi-metric evaluation, a lead ML model is selected: a tuned extra trees algorithm with performances of AUROC: 0.857 (±0.014), recall: 0.587 (±0.063), precision: 0.800 (±0.053), and f1: 0.666 (±0.042). This model can be used by the pharmaceutical industry during drug development and could even be adapted for use in clinical settings.

Predicting Drug-Target Interactions With Multi-Label Classification and Label Partitioning

Identifying drug-target interactions is crucial for drug discovery. Despite modern technologies used in drug screening, experimental identification of drug-target interactions is an extremely demanding task. Predicting drug-target interactions in silico can thereby facilitate drug discovery as well as drug repositioning. Various machine learning models have been developed over the years to predict such interactions. Multi-output learning models in particular have drawn the attention of the scientific community due to their high predictive performance and computational efficiency. These models are based on the assumption that all the labels are correlated with each other. However, this assumption is too optimistic. Here, we address drug-target interaction prediction as a multi-label classification task that is combined with label partitioning. We show that building multi-output learning models over groups (clusters) of labels often leads to superior results. The performed experiments confirm the efficiency of the proposed framework.

Time-Based Systems Biology Approaches to Capture and Model Dynamic Gene Regulatory Networks

All aspects of transcription and its regulation involve dynamic events. However, capturing these dynamic events in gene regulatory networks (GRNs) offers both a promise and a challenge. The promise is that capturing and modeling the dynamic changes in GRNs will allow us to understand how organisms adapt to a changing environment. The ability to mount a rapid transcriptional response to environmental changes is especially important in nonmotile organisms such as plants. The challenge is to capture these dynamic, genome-wide events and model them in GRNs. In this review, we cover recent progress in capturing dynamic interactions of transcription factors with their targets-at both the local and genome-wide levels-and how they are used to learn how GRNs operate as a function of time. We also discuss recent advances that employ time-based machine learning approaches to forecast gene expression at future time points, a key goal of systems biology.

Cold-Start Problems in Data-Driven Prediction of Drug-Drug Interaction Effects

Combining drugs, a phenomenon often referred to as polypharmacy, can induce additional adverse effects. The identification of adverse combinations is a key task in pharmacovigilance. In this context, in silico approaches based on machine learning are promising as they can learn from a limited number of combinations to predict for all. In this work, we identify various subtasks in predicting effects caused by drug–drug interaction. Predicting drug–drug interaction effects for drugs that already exist is very different from predicting outcomes for newly developed drugs, commonly called a cold-start problem. We propose suitable validation schemes for the different subtasks that emerge. These validation schemes are critical to correctly assess the performance. We develop a new model that obtains AUC-ROC =0.843 for the hardest cold-start task up to AUC-ROC =0.957 for the easiest one on the benchmark dataset of Zitnik et al. Finally, we illustrate how our predictions can be used to improve post-market surveillance systems or detect drug–drug interaction effects earlier during drug development.

An Artificial Neural Network Approach and a Data Augmentation Algorithm to Systematize the Diagnosis of Deep-Vein Thrombosis by Using Wells’ Criteria

The use of a back-propagation artificial neural network (ANN) to systematize the reliability of a Deep Vein Thrombosis (DVT) diagnostic by using Wells’ criteria is introduced herein. In this paper, a new ANN model is proposed to improve the Accuracy when dealing with a highly unbalanced dataset. To create the training dataset, a new data augmentation algorithm based on statistical data known as the prevalence of DVT of real cases reported in literature and from the public hospital is proposed. The above is used to generate one dataset of 10,000 synthetic cases. Each synthetic case has nine risk factors according to Wells’ criteria and also the use of two additional factors, such as gender and age, is proposed. According to interviews with medical specialists, a training scheme was established. In addition, a new algorithm is presented to improve the Accuracy and Sensitivity/Recall. According to the proposed algorithm, two thresholds of decision were found, the first one is 0.484, which is to improve Accuracy. The other one is 0.138 to improve Sensitivity/Recall. The Accuracy achieved is 90.99%, which is greater than that obtained with other related machine learning methods. The proposed ANN model was validated performing the k-fold cross validation technique using a dataset with 10,000 synthetic cases. The test was performed by using 59 real cases obtained from a regional hospital, achieving an Accuracy of 98.30%.

Network inference in systems biology: recent developments, challenges, and applications

One of the most interesting, difficult, and potentially useful topics in computational biology is the inference of gene regulatory networks (GRNs) from expression data. Although researchers have been working on this topic for more than a decade and much progress has been made, it remains an unsolved problem and even the most sophisticated inference algorithms are far from perfect. In this paper, we review the latest developments in network inference, including state-of-the-art algorithms like PIDC, Phixer, and more. We also discuss unsolved computational challenges, including the optimal combination of algorithms, integration of multiple data sources, and pseudo-temporal ordering of static expression data. Lastly, we discuss some exciting applications of network inference in cancer research, and provide a list of useful software tools for researchers hoping to conduct their own network inference analyses.

Application of deep learning methods in biological networks

The increase in biological data and the formation of various biomolecule interaction databases enable us to obtain diverse biological networks. These biological networks provide a wealth of raw materials for further understanding of biological systems, the discovery of complex diseases and the search for therapeutic drugs. However, the increase in data also increases the difficulty of biological networks analysis. Therefore, algorithms that can handle large, heterogeneous and complex data are needed to better analyze the data of these network structures and mine their useful information. Deep learning is a branch of machine learning that extracts more abstract features from a larger set of training data. Through the establishment of an artificial neural network with a network hierarchy structure, deep learning can extract and screen the input information layer by layer and has representation learning ability. The improved deep learning algorithm can be used to process complex and heterogeneous graph data structures and is increasingly being applied to the mining of network data information. In this paper, we first introduce the used network data deep learning models. After words, we summarize the application of deep learning on biological networks. Finally, we discuss the future development prospects of this field.

Prediction of the miRNA interactome - Established methods and upcoming perspectives

MicroRNAs (miRNAs) are well-studied small noncoding RNAs involved in post-transcriptional gene regulation in a wide range of organisms, including mammals. Their function is mediated by base pairing with their target RNAs. Although many features required for miRNA-mediated repression have been described, the identification of functional interactions is still challenging. In the last two decades, numerous Machine Learning (ML) models have been developed to predict their putative targets. In this review, we summarize the biological knowledge and the experimental data used to develop these ML models. Recently, Deep Neural Network-based models have also emerged in miRNA interaction modeling. We thus outline established and emerging models to give a perspective on the future developments needed to improve the identification of genes directly regulated by miRNAs.

Guiding Mineralization Co-Culture Discovery Using Bayesian Optimization

Many disciplines rely on testing combinations of compounds, materials, proteins, or bacterial species to drive scientific discovery. It is time-consuming and expensive to determine experimentally, via trial-and-error or random selection approaches, which of the many possible combinations will lead to desirable outcomes. Hence, there is a pressing need for more rational and efficient experimental design approaches to reduce experimental effort. In this work, we demonstrate the potential of machine learning methods for the in silico selection of promising co-culture combinations in the application of bioaugmentation. We use the example of pollutant removal in drinking water treatment plants, which can be achieved using co-cultures of a specialized pollutant degrader with combinations of bacterial isolates. To reduce the experimental effort needed to discover high-performing combinations, we propose a data-driven experimental design. Based on a dataset of mineralization performance for all pairs of 13 bacterial species co-cultured with MSH1, we built a Gaussian process regression model to predict the Gompertz mineralization parameters of the co-cultures of two and three species, based on the single-strain parameters. We subsequently used this model in a Bayesian optimization scheme to suggest potentially high-performing combinations of bacteria. We achieved good performance with this approach, both for predicting mineralization parameters and for selecting effective co-cultures, despite the limited dataset. As a novel application of Bayesian optimization in bioremediation, this experimental design approach has promising applications for highlighting co-culture combinations for in vitro testing in various settings, to lessen the experimental burden and perform more targeted screenings.

Network inference with ensembles of bi-clustering trees

BACKGROUND

Network inference is crucial for biomedicine and systems biology. Biological entities and their associations are often modeled as interaction networks. Examples include drug protein interaction or gene regulatory networks. Studying and elucidating such networks can lead to the comprehension of complex biological processes. However, usually we have only partial knowledge of those networks and the experimental identification of all the existing associations between biological entities is very time consuming and particularly expensive. Many computational approaches have been proposed over the years for network inference, nonetheless, efficiency and accuracy are still persisting open problems. Here, we propose bi-clustering tree ensembles as a new machine learning method for network inference, extending the traditional tree-ensemble models to the global network setting. The proposed approach addresses the network inference problem as a multi-label classification task. More specifically, the nodes of a network (e.g., drugs or proteins in a drug-protein interaction network) are modelled as samples described by features (e.g., chemical structure similarities or protein sequence similarities). The labels in our setting represent the presence or absence of links connecting the nodes of the interaction network (e.g., drug-protein interactions in a drug-protein interaction network).

RESULTS

We extended traditional tree-ensemble methods, such as extremely randomized trees (ERT) and random forests (RF) to ensembles of bi-clustering trees, integrating background information from both node sets of a heterogeneous network into the same learning framework. We performed an empirical evaluation, comparing the proposed approach to currently used tree-ensemble based approaches as well as other approaches from the literature. We demonstrated the effectiveness of our approach in different interaction prediction (network inference) settings. For evaluation purposes, we used several benchmark datasets that represent drug-protein and gene regulatory networks. We also applied our proposed method to two versions of a chemical-protein association network extracted from the STITCH database, demonstrating the potential of our model in predicting non-reported interactions.

CONCLUSIONS

Bi-clustering trees outperform existing tree-based strategies as well as machine learning methods based on other algorithms. Since our approach is based on tree-ensembles it inherits the advantages of tree-ensemble learning, such as handling of missing values, scalability and interpretability.

Fast Kronecker Product Kernel Methods via Generalized Vec Trick

Kronecker product kernel provides the standard approach in the kernel methods' literature for learning from graph data, where edges are labeled and both start and end vertices have their own feature representations. The methods allow generalization to such new edges, whose start and end vertices do not appear in the training data, a setting known as zero-shot or zero-data learning. Such a setting occurs in numerous applications, including drug-target interaction prediction, collaborative filtering, and information retrieval. Efficient training algorithms based on the so-called vec trick that makes use of the special structure of the Kronecker product are known for the case where the training data are a complete bipartite graph. In this paper, we generalize these results to noncomplete training graphs. This allows us to derive a general framework for training Kronecker product kernel methods, as specific examples we implement Kronecker ridge regression and support vector machine algorithms. Experimental results demonstrate that the proposed approach leads to accurate models, while allowing order of magnitude improvements in training and prediction time.

Ranking genome-wide correlation measurements improves microarray and RNA-seq based global and targeted co-expression networks

Co-expression networks are essential tools to infer biological associations between gene products and predict gene annotation. Global networks can be analyzed at the transcriptome-wide scale or after querying them with a set of guide genes to capture the transcriptional landscape of a given pathway in a process named Pathway Level Coexpression (PLC). A critical step in network construction remains the definition of gene co-expression. In the present work, we compared how Pearson Correlation Coefficient (PCC), Spearman Correlation Coefficient (SCC), their respective ranked values (Highest Reciprocal Rank (HRR)), Mutual Information (MI) and Partial Correlations (PC) performed on global networks and PLCs. This evaluation was conducted on the model plant Arabidopsis thaliana using microarray and differently pre-processed RNA-seq datasets. We particularly evaluated how dataset × distance measurement combinations performed in 5 PLCs corresponding to 4 well described plant metabolic pathways (phenylpropanoid, carbohydrate, fatty acid and terpene metabolisms) and the cytokinin signaling pathway. Our present work highlights how PCC ranked with HRR is better suited for global network construction and PLC with microarray and RNA-seq data than other distance methods, especially to cluster genes in partitions similar to biological subpathways.

A Comparative Study of Pairwise Learning Methods Based on Kernel Ridge Regression

Many machine learning problems can be formulated as predicting labels for a pair of objects. Problems of that kind are often referred to as pairwise learning, dyadic prediction, or network inference problems. During the past decade, kernel methods have played a dominant role in pairwise learning. They still obtain a state-of-the-art predictive performance, but a theoretical analysis of their behavior has been underexplored in the machine learning literature. In this work we review and unify kernel-based algorithms that are commonly used in different pairwise learning settings, ranging from matrix filtering to zero-shot learning. To this end, we focus on closed-form efficient instantiations of Kronecker kernel ridge regression. We show that independent task kernel ridge regression, two-step kernel ridge regression, and a linear matrix filter arise naturally as a special case of Kronecker kernel ridge regression, implying that all these methods implicitly minimize a squared loss. In addition, we analyze universality, consistency, and spectral filtering properties. Our theoretical results provide valuable insights into assessing the advantages and limitations of existing pairwise learning methods.

Connectivity inference from neural recording data: Challenges, mathematical bases and research directions

This article presents a review of computational methods for connectivity inference from neural activity data derived from multi-electrode recordings or fluorescence imaging. We first identify biophysical and technical challenges in connectivity inference along the data processing pipeline. We then review connectivity inference methods based on two major mathematical foundations, namely, descriptive model-free approaches and generative model-based approaches. We investigate representative studies in both categories and clarify which challenges have been addressed by which method. We further identify critical open issues and possible research directions.

A Machine Learning Approach to Predict Gene Regulatory Networks in Seed Development in Arabidopsis

Gene regulatory networks (GRNs) provide a representation of relationships between regulators and their target genes. Several methods for GRN inference, both unsupervised and supervised, have been developed to date. Because regulatory relationships consistently reprogram in diverse tissues or under different conditions, GRNs inferred without specific biological contexts are of limited applicability. In this report, a machine learning approach is presented to predict GRNs specific to developing Arabidopsis thaliana embryos. We developed the Beacon GRN inference tool to predict GRNs occurring during seed development in Arabidopsis based on a support vector machine (SVM) model. We developed both global and local inference models and compared their performance, demonstrating that local models are generally superior for our application. Using both the expression levels of the genes expressed in developing embryos and prior known regulatory relationships, GRNs were predicted for specific embryonic developmental stages. The targets that are strongly positively correlated with their regulators are mostly expressed at the beginning of seed development. Potential direct targets were identified based on a match between the promoter regions of these inferred targets and the cis elements recognized by specific regulators. Our analysis also provides evidence for previously unknown inhibitory effects of three positive regulators of gene expression. The Beacon GRN inference tool provides a valuable model system for context-specific GRN inference and is freely available at https://github.com/BeaconProjectAtVirginiaTech/beacon_network_inference.git.

Computational inference of gene regulatory networks: Approaches, limitations and opportunities

Gene regulatory networks lie at the core of cell function control. In E. coli and S. cerevisiae, the study of gene regulatory networks has led to the discovery of regulatory mechanisms responsible for the control of cell growth, differentiation and responses to environmental stimuli. In plants, computational rendering of gene regulatory networks is gaining momentum, thanks to the recent availability of high-quality genomes and transcriptomes and development of computational network inference approaches. Here, we review current techniques, challenges and trends in gene regulatory network inference and highlight challenges and opportunities for plant science. We provide plant-specific application examples to guide researchers in selecting methodologies that suit their particular research questions. Given the interdisciplinary nature of gene regulatory network inference, we tried to cater to both biologists and computer scientists to help them engage in a dialogue about concepts and caveats in network inference. Specifically, we discuss problems and opportunities in heterogeneous data integration for eukaryotic organisms and common caveats to be considered during network model evaluation. This article is part of a Special Issue entitled: Plant Gene Regulatory Mechanisms and Networks, edited by Dr. Erich Grotewold and Dr. Nathan Springer.

Integration of omic networks in a developmental atlas of maize

Coexpression networks and gene regulatory networks (GRNs) are emerging as important tools for predicting functional roles of individual genes at a system-wide scale. To enable network reconstructions, we built a large-scale gene expression atlas composed of 62,547 messenger RNAs (mRNAs), 17,862 nonmodified proteins, and 6227 phosphoproteins harboring 31,595 phosphorylation sites quantified across maize development. Networks in which nodes are genes connected on the basis of highly correlated expression patterns of mRNAs were very different from networks that were based on coexpression of proteins. Roughly 85% of highly interconnected hubs were not conserved in expression between RNA and protein networks. However, networks from either data type were enriched in similar ontological categories and were effective in predicting known regulatory relationships. Integration of mRNA, protein, and phosphoprotein data sets greatly improved the predictive power of GRNs.

Machine Learning of Protein Interactions in Fungal Secretory Pathways

In this paper we apply machine learning methods for predicting protein interactions in fungal secretion pathways. We assume an inter-species transfer setting, where training data is obtained from a single species and the objective is to predict protein interactions in other, related species. In our methodology, we combine several state of the art machine learning approaches, namely, multiple kernel learning (MKL), pairwise kernels and kernelized structured output prediction in the supervised graph inference framework. For MKL, we apply recently proposed centered kernel alignment and p-norm path following approaches to integrate several feature sets describing the proteins, demonstrating improved performance. For graph inference, we apply input-output kernel regression (IOKR) in supervised and semi-supervised modes as well as output kernel trees (OK3). In our experiments simulating increasing genetic distance, Input-Output Kernel Regression proved to be the most robust prediction approach. We also show that the MKL approaches improve the predictions compared to uniform combination of the kernels. We evaluate the methods on the task of predicting protein-protein-interactions in the secretion pathways in fungi, S.cerevisiae, baker's yeast, being the source, T. reesei being the target of the inter-species transfer learning. We identify completely novel candidate secretion proteins conserved in filamentous fungi. These proteins could contribute to their unique secretion capabilities.

Prediction of Genetic Interactions Using Machine Learning and Network Properties

A genetic interaction (GI) is a type of interaction where the effect of one gene is modified by the effect of one or several other genes. These interactions are important for delineating functional relationships among genes and their corresponding proteins, as well as elucidating complex biological processes and diseases. An important type of GI - synthetic sickness or synthetic lethality - involves two or more genes, where the loss of either gene alone has little impact on cell viability, but the combined loss of all genes leads to a severe decrease in fitness (sickness) or cell death (lethality). The identification of GIs is an important problem for it can help delineate pathways, protein complexes, and regulatory dependencies. Synthetic lethal interactions have important clinical and biological significance, such as providing therapeutically exploitable weaknesses in tumors. While near systematic high-content screening for GIs is possible in single cell organisms such as yeast, the systematic discovery of GIs is extremely difficult in mammalian cells. Therefore, there is a great need for computational approaches to reliably predict GIs, including synthetic lethal interactions, in these organisms. Here, we review the state-of-the-art approaches, strategies, and rigorous evaluation methods for learning and predicting GIs, both under general (healthy/standard laboratory) conditions and under specific contexts, such as diseases.

Classifying pairs with trees for supervised biological network inference

Networks are ubiquitous in biology, and computational approaches have been largely investigated for their inference. In particular, supervised machine learning methods can be used to complete a partially known network by integrating various measurements. Two main supervised frameworks have been proposed: the local approach, which trains a separate model for each network node, and the global approach, which trains a single model over pairs of nodes. Here, we systematically investigate, theoretically and empirically, the exploitation of tree-based ensemble methods in the context of these two approaches for biological network inference. We first formalize the problem of network inference as a classification of pairs, unifying in the process homogeneous and bipartite graphs and discussing two main sampling schemes. We then present the global and the local approaches, extending the latter for the prediction of interactions between two unseen network nodes, and discuss their specializations to tree-based ensemble methods, highlighting their interpretability and drawing links with clustering techniques. Extensive computational experiments are carried out with these methods on various biological networks that clearly highlight that these methods are competitive with existing methods.

Network Assessor: an automated method for quantitative assessment of a network's potential for gene function prediction

Significant effort has been invested in network-based gene function prediction algorithms based on the guilt by association (GBA) principle. Existing approaches for assessing prediction performance typically compute evaluation metrics, either averaged across all functions being considered, or strictly from properties of the network. Since the success of GBA algorithms depends on the specific function being predicted, evaluation metrics should instead be computed for each function. We describe a novel method for computing the usefulness of a network by measuring its impact on gene function cross validation prediction performance across all gene functions. We have implemented this in software called Network Assessor, and describe its use in the GeneMANIA (GM) quality control system. Network Assessor is part of the GM command line tools.