Computational Prediction Methods Research Articles

Computational insight into the three-dimensional structure of ADP ribosylation factor like protein 15, a novel susceptibility gene for rheumatoid arthritis

The ARL15 gene (ADP ribosylation factor like protein 15) encodes for an uncharacterized small GTP-binding protein. Its exact role in human physiology remains unknown, but a number of genetic association studies have recognised different variants in this gene to be statistically associated with numerous traits and complex diseases. We have previously reported a novel association of ARL15 with rheumatoid arthritis (RA) based on a genome-wide association study in a north Indian cohort. Subsequent investigations have provided leads for its involvement in RA pathophysiology, especially its potential as a novel therapeutic target. However, the absence of an experimentally determined tertiary structure for ARL15 significantly hinders the understanding of its biochemical and physiological functions, as well as development of potential lead molecules. We, therefore, aimed to derive a high quality, refined model of the three dimensional structure of human ARL15 protein using two different computational protein structure prediction methods – template-based threading and ab initio modelling. The best model each from among the five each derived from both the approaches was selected based on stringent quality assessment and refinement. Molecular dynamics simulations over long timescales revealed the ab initio model to be relatively more stable, and it marginally outperformed the template-based model in the quality assessment as well. A putative GTP-binding site was also predicted using homology for the ARL15 protein, where potential competitive inhibitors can be targeted. This high quality predicted model may provide insights to the biological role(s) of ARL15 and inform and guide further experimental, structural and biochemical characterization efforts. Communicated by Ramaswamy H. Sarma

High-order three-scale computational method for elastic behavior analysis and strength prediction of axisymmetric composite structures with multiple spatial scales

A novel high-order three-scale (HOTS) computational method for elastic behavior analysis and strength prediction of axisymmetric composite structures with multiple spatial scales is developed in this paper. The multiple heterogeneities of axisymmetric composite structures we investigated are taken into account by periodic distributions of representative unit cells on the mesoscale and microscale. First, the new micro–meso–macro coupled HOTS computational model for elastic problems of axisymmetric composite structures is established based on multiscale asymptotic analysis, which breaks through the traditional multiscale assumptions and includes three main components. Two classes of mesoscopic and microscopic auxiliary cell functions are constructed on the mesoscale and microscale, respectively. The macroscopic homogenization problems are defined on global axisymmetric structures by twice up-scaling procedures from microscale to mesoscale and then from mesoscale to macroscale. Moreover, the asymptotic HOTS solutions are constructed for approximating multiscale elastic problems of axisymmetric structures and the numerical accuracy analysis of the HOTS solutions is given in detail. Then, the strength prediction formulas for axisymmetric composite structures with multiple spatial scales are presented based on the high-accuracy elastic behavior analysis from the proposed HOTS computational model. Furthermore, the corresponding HOTS numerical algorithm based on the finite element method (FEM) is presented for analyzing the mechanical behaviors and predicting the strength of axisymmetric composite structures with multiple spatial scales in detail. Finally, some numerical examples are reported to verify the feasibility and effectiveness of the proposed HOTS computational method. In this study, a unified three-scale computational framework is offered, which enables the simulation of mechanical behaviors of axisymmetric composite structures with multiple spatial scales.

Predicting Long non-coding RNAs through feature ensemble learning

BackgroundMany transcripts have been generated due to the development of sequencing technologies, and lncRNA is an important type of transcript. Predicting lncRNAs from transcripts is a challenging and important task. Traditional experimental lncRNA prediction methods are time-consuming and labor-intensive. Efficient computational methods for lncRNA prediction are in demand.ResultsIn this paper, we propose two lncRNA prediction methods based on feature ensemble learning strategies named LncPred-IEL and LncPred-ANEL. Specifically, we encode sequences into six different types of features including transcript-specified features and general sequence-derived features. Then we consider two feature ensemble strategies to utilize and integrate the information in different feature types, the iterative ensemble learning (IEL) and the attention network ensemble learning (ANEL). IEL employs a supervised iterative way to ensemble base predictors built on six different types of features. ANEL introduces an attention mechanism-based deep learning model to ensemble features by adaptively learning the weight of individual feature types. Experiments demonstrate that both LncPred-IEL and LncPred-ANEL can effectively separate lncRNAs and other transcripts in feature space. Moreover, comparison experiments demonstrate that LncPred-IEL and LncPred-ANEL outperform several state-of-the-art methods when evaluated by 5-fold cross-validation. Both methods have good performances in cross-species lncRNA prediction.ConclusionsLncPred-IEL and LncPred-ANEL are promising lncRNA prediction tools that can effectively utilize and integrate the information in different types of features.

MiRTMC: A miRNA Target Prediction Method Based on Matrix Completion Algorithm.

microRNAs (miRNAs) are small non-coding RNAs which modulate the stability of gene targets and their rates of translation into proteins at transcriptional level and post-transcriptional level. miRNA dysfunctions can lead to human diseases because of dysregulation of their targets. Correct miRNA target prediction will lead to better understanding of the mechanisms of human diseases and provide hints on curing them. In recent years, computational miRNA target prediction methods have been proposed according to the interaction rules between miRNAs and targets. However, these methods suffer from high false positive rates due to the complicated relationship between miRNAs and their targets. The rapidly growing number of experimentally validated miRNA targets enables predicting miRNA targets with high precision via accurate data analysis. Taking advantage of these known miRNA targets, a novel recommendation system model (miRTMC) for miRNA target prediction is established using a new matrix completion algorithm. In miRTMC, a heterogeneous network is constructed by integrating the miRNA similarity network, the gene similarity network, and the miRNA-gene interaction network. Our assumption is that the latent factors determining whether a gene is the target of miRNA or not are highly correlated, i.e., the adjacency matrix of the heterogeneous network is low-rank, which is then completed by using a nuclear norm regularized linear least squares model under non-negative constraints. Alternating direction method of multipliers (ADMM) is adopted to numerically solve the matrix completion problem. Our results show that miRTMC outperforms the competing methods in terms of various evaluation metrics. Our software package is available at https://github.com/hjiangcsu/miRTMC.

Cellular sociology regulates the hierarchical spatial patterning and organization of cells in organisms.

Advances in single-cell biotechnology have increasingly revealed interactions of cells with their surroundings, suggesting a cellular society at the microscale. Similarities between cells and humans across multiple hierarchical levels have quantitative inference potential for reaching insights about phenotypic interactions that lead to morphological forms across multiple scales of cellular organization, namely cells, tissues and organs. Here, the functional and structural comparisons between how cells and individuals fundamentally socialize to give rise to the spatial organization are investigated. Integrative experimental cell interaction assays and computational predictive methods shape the understanding of societal perspective in the determination of the cellular interactions that create spatially coordinated forms in biological systems. Emerging quantifiable models from a simpler biological microworld such as bacterial interactions and single-cell organisms are explored, providing a route to model spatio-temporal patterning of morphological structures in humans. This analogical reasoning framework sheds light on structural patterning principles as a result of biological interactions across the cellular scale and up.

Computer Assessment of the Xenobiotic Metabolites Formation’s Probability in the Human Body

Abstract—Xenobiotics undergo biotransformation in the human body and the resulting metabolites may greatly differ in their physical, chemical, and biological properties from the initial substances or be toxic or reactive. An experimental study of xenobiotic biotransformation is challenging; designing computational prediction methods is therefore important. One major drawback of computational methods is that a large number of possible metabolites are generated, leading to a combinatorial explosion. The objective of this work was to select the criteria for optimizing the prediction of metabolites via the original web resource MetaTox ( http://www.way2drug.com/MG ), which is publicly available. The performance was compared for additive and multiplicative methods to assess the probability of metabolite formation; the additive approach was found to be superior.

The assessment of three dimensional modelling design for single strand DNA aptamers for computational chemistry application

Aptamers are oligonucleotides and peptides around 15–100 bases in length and are suitable as detection probes or as therapeutics molecules. There are growing interests in the aptamer screening approach through computational simulation methods. DNA and RNA modelling lacks of validation on their predicted 3D structures due to less number of validation tools, unlike protein structures. We suggest an approach to design the stem-loop/hairpin for the three dimensional structure of DNA aptamers through serial applications of computational prediction methods by comparing the simulated structures with the experimental data deposited in PDB Data bank, followed by MD simulations. The result shows minimal structural differences were observed between the designed and the original NMR aptamers, and the stem-loop conformational structures were also retained during the MD thus suggesting the proposed aptamers designing methods are able to synthesize a high quality molecular structure of hairpin aptamers, comparable to the NMR structures.

DeepAdd: Protein function prediction from k-mer embedding and additional features

With the application of new high throughput sequencing technology, a large number of protein sequences is becoming available. Determination of the functional characteristics of these proteins by experiments is an expensive endeavor that requires a lot of time. Furthermore, at the organismal level, such kind of experimental functional analyses can be conducted only for a very few selected model organisms. Computational function prediction methods can be used to fill this gap. The functions of proteins are classified by Gene Ontology (GO), which contains more than 40,000 classifications in three domains, Molecular Function (MF), Biological Process (BP), and Cellular Component (CC). Additionally, since proteins have many functions, function prediction represents a multi-label and multi-class problem. We developed a new method to predict protein function from sequence. To this end, natural language model was used to generate word embedding of sequence and learn features from it by deep learning, and additional features to locate every protein. Our method uses the dependencies between GO classes as background information to construct a deep learning model. We evaluate our method using the standards established by the Computational Assessment of Function Annotation (CAFA) and have noticeable improvement over several algorithms, such as FFPred, DeepGO, GoFDR and other methods compared on the CAFA3 datasets.

Transcriptional enhancers: from prediction to functional assessment on a genome-wide scale.

Enhancers are cis-regulatory sequences located distally to target genes. These sequences consolidate developmental and environmental cues to coordinate gene expression in a tissue-specific manner. Enhancer function and tissue specificity depend on the expressed set of transcription factors, which recognize binding sites and recruit cofactors that regulate local chromatin organization and gene transcription. Unlike other genomic elements, enhancers are challenging to identify because they function independently of orientation, are often distant from their promoters, have poorly defined boundaries, and display no reading frame. In addition, there are no defined genetic or epigenetic features that are unambiguously associated with enhancer activity. Over recent years there have been developments in both empirical assays and computational methods for enhancer prediction. We review genome-wide tools, CRISPR advancements, and high-throughput screening approaches that have improved our ability to both observe and manipulate enhancers in vitro at the level of primary genetic sequences, chromatin states, and spatial interactions. We also highlight contemporary animal models and their importance to enhancer validation. Together, these experimental systems and techniques complement one another and broaden our understanding of enhancer function in development, evolution, and disease.

Identification of small non-coding RNAs from Rhizobium etli by integrated genome-wide and transcriptome-based methods

BackgroundSmall non-coding RNAs (sRNAs) are regulatory molecules, present in all forms of life, known to regulate various biological processes in response to the different environmental signals. In recent years, deep sequencing and various other computational prediction methods have been employed to identify and analyze sRNAs.ResultsIn the present study, we have applied an improved sRNA scanner method to predict sRNAs from the genome of Rhizobium etli, based on PWM matrix of conditional sigma factor 32. sRNAs predicted from the genome are integrated with the available stress specific transcriptome data to predict putative conditional specific sRNAs. A total of 271 sRNAs from the genome and 173 sRNAs from the transcriptome are computationally predicted. Of these, 25 sRNAs are found in both genome and transcriptome data. Putative targets for these sRNAs are predicted using TargetRNA2 and these targets are involved in a wide array of cellular functions such as cell division, transport and metabolism of amino acids, carbohydrates, energy production and conversion, translation, cell wall/membrane biogenesis, post-translation modification, protein turnover and chaperones. Predicted targets are functionally classified based on COG analysis and GO annotations.ConclusionsRNAs predicted from the genome, using PWM matrices for conditional sigma factor 32 could be a better method to identify the conditional specific sRNAs which expand the list of putative sRNAs from the intergenic regions (IgRs) of R. etli and closely related α-proteobacteria. sRNAs identified in this study would be helpful to explore their regulatory role in biological cellular process during the stress.

MiRNALoc: predicting miRNA subcellular localizations based on principal component scores of physico-chemical properties and pseudo compositions of di-nucleotides

MicroRNAs (miRNAs) are one kind of non-coding RNA, play vital role in regulating several physiological and developmental processes. Subcellular localization of miRNAs and their abundance in the native cell are central for maintaining physiological homeostasis. Besides, RNA silencing activity of miRNAs is also influenced by their localization and stability. Thus, development of computational method for subcellular localization prediction of miRNAs is desired. In this work, we have proposed a computational method for predicting subcellular localizations of miRNAs based on principal component scores of thermodynamic, structural properties and pseudo compositions of di-nucleotides. Prediction accuracy was analyzed following fivefold cross validation, where ~ 63–71% of AUC-ROC and ~ 69–76% of AUC-PR were observed. While evaluated with independent test set, > 50% localizations were found to be correctly predicted. Besides, the developed computational model achieved higher accuracy than the existing methods. A user-friendly prediction server “miRNALoc” is freely accessible at http://cabgrid.res.in:8080/mirnaloc/, by which the user can predict localizations of miRNAs.

CAPICE: a computational method for Consequence-Agnostic Pathogenicity Interpretation of Clinical Exome variations

Exome sequencing is now mainstream in clinical practice. However, identification of pathogenic Mendelian variants remains time-consuming, in part, because the limited accuracy of current computational prediction methods requires manual classification by experts. Here we introduce CAPICE, a new machine-learning-based method for prioritizing pathogenic variants, including SNVs and short InDels. CAPICE outperforms the best general (CADD, GAVIN) and consequence-type-specific (REVEL, ClinPred) computational prediction methods, for both rare and ultra-rare variants. CAPICE is easily added to diagnostic pipelines as pre-computed score file or command-line software, or using online MOLGENIS web service with API. Download CAPICE for free and open-source (LGPLv3) at https://github.com/molgenis/capice.

A multiclass SVM classifier with teaching learning based feature subset selection for enzyme subclass classification

The functional characterization of proteins aims to understand life at macroscopic and microscopic levels, thus having wide and extensive applications in biological and pharmaceutical research. The classical methods used for the functional characterization of proteins rely on the sequence similarity approach i.e. identifying a similar protein whose functions are known. However, the performance of these methods reduces in the event of the sequences exhibiting less similarity. Hence, to tide over this impediment, many computational function prediction methods that do not use the sequence similarity approach were developed. This paper focuses on a Support Vector Machine (SVM) based method which classifies a protein at three levels. At the first level, a protein is identified and earmarked as an enzyme or non-enzyme. In the case of it being an enzyme, its functional class, and subclass are predicted in one step by taking 7 EC classes with 63 of their subclasses. Each protein is initially represented by its 32 physicochemical properties. Then a multiclass SVM (MSVM) classifier, that solves an n class classification problem with log2n binary classifiers, along with a Modified Teaching Learning Based Optimization method (MTLBO) to identify the significant features for classification, is designed to predict the functional class, and subclass of an enzyme in one step. The recall of the proposed MSVM-MTLBO using only 25 features ranges from 92.97% to 98.14% for class label prediction and 60% to 98.25% for subclass label prediction, which is significantly better than the existing methods for class and subclass classification. Moreover, the proposed method predicts the subclass labels with less number of binary classifiers than a standard multiclass SVM.

Variational graph auto-encoders for miRNA-disease association prediction.

Cumulative experimental studies have demonstrated the critical roles of microRNAs (miRNAs) in the diverse fundamental and important biological processes, and in the development of numerous complex human diseases. Thus, exploring the relationships between miRNAs and diseases is helpful with understanding the mechanisms, the detection, diagnosis, and treatment of complex diseases. As the identification of miRNA-disease associations via traditional biological experiments is time-consuming and expensive, an effective computational prediction method is appealing. In this study, we present a deep learning framework with variational graph auto-encoder for miRNA-disease association prediction (VGAE-MDA). VGAE-MDA first gets the representations of miRNAs and diseases from the heterogeneous networks constructed by miRNA-miRNA similarity, disease-disease similarity, and known miRNA-disease associations. Then, VGAE-MDA constructs two sub-networks: miRNA-based network and disease-based network. Combining the representations based on the heterogeneous network, two variational graph auto-encoders (VGAE) are deployed for calculating the miRNA-disease association scores from two sub-networks, respectively. Lastly, VGAE-MDA obtains the final predicted association score for a miRNA-disease pair by integrating the scores from these two trained networks. Unlike the previous model, the VGAE-MDA can mitigate the effect of noises from random selection of negative samples. Besides, the use of graph convolutional neural (GCN) network can naturally incorporate the node features from the graph structure while the variational autoencoder (VAE) makes use of latent variables to predict associations from the perspective of data distribution. The experimental results show that VGAE-MDA outperforms the state-of-the-art approaches in miRNA-disease association prediction. Besides, the effectiveness of our model has been further demonstrated by case studies.

Evaluation of Free Energy Calculations for the Prioritization of Macrocycle Synthesis

A tremendous research and development effort was exerted toward combating chronic hepatitis C, ultimately leading to curative oral treatments, all of which are targeting viral proteins. Despite the advantage of numerous targets allowing for broad hepatitis C virus (HCV) genotype coverage, the only host target inhibitors that advanced into clinical development were Cyclosporin A based cyclophilin inhibitors. While cyclosporin-based molecules typically require a fermentation process, Gilead successfully pursued a fully synthetic, oral program based on Sanglifehrin A. The drug discovery process, though greatly helped by facile crystallography, was still hampered by the limitations in the accuracy of predictive computational methods for prioritizing compound ideas. Recent advances in accuracy and speed of free energy perturbation (FEP) methods, however, are attractive for prioritizing and derisking synthetically challenging molecules and potentially could have had a significant impact on the speed of the development of this program. Here in our simulated prospective study, the binding free energies of 26 macrocyclic cyclophilin inhibitors were blindly predicted using FEP+ to test this hypothesis. The predictions had a low mean unsigned error (MUE) (1.1 kcal/mol) and accurately reproduced many design decisions from the program, suggesting that FEP+ has the potential to drive synthetic chemistry efforts by more accurately ranking compounds with nonintuitive structure-activity relationships (SARs).

Ab Initio Prediction of Metal-Organic Framework Structures

Metal-organic frameworks (MOFs) have emerged as highly versatile materials with applications in gas storage and separation, solar light energy harvesting and photocatalysis. The design of new MOFs, however, has been hampered by the lack of computational methods for ab initio MOF structure prediction, which could be used to inspire and direct experimental synthesis. Here, we report the first ab initio method for the prediction of MOF structures and test it against a diverse set of known MOFs that were chosen for their differences in topology, metal coordination geometry, and ligand binding sites. In all cases, our calculations produced structures that match experiment using only the target composition and ligand molecular structure, proving the versatility of our procedure. The herein presented methodology utilizes the point group symmetry of ligands to enable, for the first time, prediction of MOF structures from first principles, without having to resort to empirical guidelines based on rigid connectivity of nodes and linkers, or to previously determined crystal structures and topologies of known MOFs. This advance provides the first tool to change MOF design from an empirically based process that is based on chemistʼs intuition rooted in literature- or database-established knowledge of node-and-linker connectivity to a more general and theory-driven materials development. This ab initio MOF structure prediction approach, which is here validated on a range of known MOF classes, provides a unique opportunity to explore the phase landscape of MOFs computationally and enables MOF research and development even in case of limited access to laboratory resources, as for example in case of a global pandemic.

Predicting drug-drug interactions using multi-modal deep auto-encoders based network embedding and positive-unlabeled learning.

Drug-drug interactions (DDIs) are crucial for public health and patient safety, which has aroused widespread concern in academia and industry. The existing computational DDI prediction methods are mainly divided into four categories: literature extraction-based, similarity-based, matrix operations-based and network-based. A number of recent studies have revealed that integrating heterogeneous drug features is of significant importance for developing high-accuracy prediction models. Meanwhile, drugs that lack certain features could utilize other features to learn representations. However, it also brings some new challenges such as incomplete data, non-linear relations and heterogeneous properties. In this paper, we propose a multi-modal deep auto-encoders based drug representation learning method named DDI-MDAE, to predict DDIs from large-scale, noisy and sparse data. Our method aims to learn unified drug representations from multiple drug feature networks simultaneously using multi-modal deep auto-encoders. Then, we apply four operators on the learned drug embeddings to represent drug-drug pairs and adopt the random forest classifier to train models for predicting DDIs. The experimental results demonstrate the effectiveness of our proposed method for DDI prediction and significant improvement compared to other state-of-the-art benchmark methods. Moreover, we apply a specialized random forest classifier in the positive-unlabeled (PU) learning setting to enhance the prediction accuracy. Experimental results reveal that the model improved by PU learning outperforms the original method DDI-MDAE by 7.1% and 6.2% improvement in AUPR metric respectively on 3-fold cross-validation (3-CV) and 5-fold cross-validation (5-CV). And in F-measure metric, the improved model gains 10.4% and 8.4% improvement over DDI-MDAE respectively on 3-CV and 5-CV. The usefulness of DDI-MDAE is further demonstrated by case studies.

Decoding Protein-protein Interactions: An Overview.

Drug discovery has focused on the paradigm "one drug, one target" for a long time. However, small molecules can act at multiple macromolecular targets, which serves as the basis for drug repurposing. In an effort to expand the target space, and given advances in X-ray crystallography, protein-protein interactions have become an emerging focus area of drug discovery enterprises. Proteins interact with other biomolecules and it is this intricate network of interactions that determines the behavior of the system and its biological processes. In this review, we briefly discuss networks in disease, followed by computational methods for protein-protein complex prediction. Computational methodologies and techniques employed towards objectives such as protein-protein docking, protein-protein interactions, and interface predictions are described extensively. Docking aims at producing a complex between proteins, while interface predictions identify a subset of residues on one protein that could interact with a partner, and protein-protein interaction sites address whether two proteins interact. In addition, approaches to predict hot spots and binding sites are presented along with a representative example of our internal project on the chemokine CXC receptor 3 B-isoform and predictive modeling with IP10 and PF4.

Prediction of First Lactation 305-day Milk Yield Based on Bimonthly Test Day Milk Yield Records in Murrah Buffaloes

The present study was conducted on 2100 first lactation bimonthly test day milk yield (BTDY) records of 350 Murrah buffaloes calved in between 1993 and 2017 at ICAR-NDRI, Karnal. A total of 6 BTDY records were taken from each animal at an interval of 60 days, from 6th day to 305th day of lactation. The prediction of First Lactation 305-Day Milk Yield (FL305DMY) was done by utilizing five conventional and machine learning methods viz., Centering Date Method (CDM), Test Interval Method (TIM), Ratio Method (RM), Multiple Linear Regression (MLR) and Artificial Neural Network (ANN). Error in prediction was estimated by absolute error, percentage absolute error, average error, percentage average error, Root Mean Square Error (RMSE) and percentage RMSE. MLR was found to be the best method with the least error in prediction (5.71% RMSE), followed by ANN (5.77% RMSE). The accuracy (R2) of MLR equation including all 6 BTDY records was 91.86%. The best MLR equation for an early prediction of FL305DMY included 3 BTDY records viz., BTDY-2 (65th day), BTDY-3 (125th day) and BTDY-4 (185th day) with 85.29% R2. The study compared the conventional and computational methods for prediction of first lactation milk yield which could be used for early selection of the animals.

Deconvolution of conformational exchange from Raman spectra of aqueous RNA nucleosides

Ribonucleic acids (RNAs) are key to the central dogma of molecular biology. While Raman spectroscopy holds great potential for studying RNA conformational dynamics, current computational Raman prediction and assignment methods are limited in terms of system size and inclusion of conformational exchange. Here, a framework is presented that predicts Raman spectra using mixtures of sub-spectra corresponding to major conformers calculated using classical and ab initio molecular dynamics. Experimental optimization allowed purines and pyrimidines to be characterized as predominantly syn and anti, respectively, and ribose into exchange between equivalent south and north populations. These measurements are in excellent agreement with Raman spectroscopy of ribonucleosides, and previous experimental and computational results. This framework provides a measure of ribonucleoside solution populations and conformational exchange in RNA subunits. It complements other experimental techniques and could be extended to other molecules, such as proteins and carbohydrates, enabling biological insights and providing a new analytical tool.