Current Bioinformatics - Volume 16, Issue 2, 2021

Background: Epigenetic repression mechanisms play an important role in gene regulation, specifically in cancer development. In many cases, a CpG island’s (CGI) susceptibility or resistance to methylation is shown to be contributed by local DNA sequence features. Objective: To develop unbiased machine learning models–individually and combined for different biological features–that predict the methylation propensity of a CGI. Methods: We developed our model consisting of CGI sequence features on a dataset of 75 sequences (28 prone, 47 resistant) representing a genome-wide methylation structure. We tested our model on two independent datasets that are chromosome (132 sequences) and disease (70 sequences) specific. Results: We provided improvements in prediction accuracy over previous models. Our results indicate that combined features better predict the methylation propensity of a CGI (area under the curve (AUC) ~0.81). Our global methylation classifier performs well on independent datasets reaching an AUC of ~0.82 for the complete model and an AUC of ~0.88 for the model using select sequences that better represent their classes in the training set. We report certain de novo motifs and transcription factor binding site (TFBS) motifs that are consistently better in separating prone and resistant CGIs. Conclusion: Predictive models for the methylation propensity of CGIs lead to a better understanding of disease mechanisms and can be used to classify genes based on their tendency to contain methylation prone CGIs, which may lead to preventative treatment strategies. MATLAB® and Python™ scripts used for model building, prediction, and downstream analyses are available at https://github.com/dicleyalcin/methylProp_predictor.

Objective: To identify the possible virulence factors (VFs) of P. proliferus. Methods: By Illumina HiSeq 4000 RNA-Seq platform, transcriptomes of adult P. proliferus worms were sequenced to predict VFs via screening the homologues of traditional VFs of parasites based on the annotations in the functional databases. Homology analysis was also performed to screen homologous genes between P. proliferus and other four Paragonimus species (i.e., P. kellicotti, P. skrjabini, P. miyazakii and P. westermani) whose transcriptomes were downloaded from the National Center for Biotechnology Information (NCBI) database, and then the differential-expressed homologous genes (DEHGs) were screened via comparisons of P. proliferus and P. kellicotti, P. skrjabini, P. miyazakii and P. westermani, respectively. Finally, an overlap of the predicted VFs and DEHGs was performed to identify possible key VFs that do not only belong to the predicted VFs but also DEHGs. Results: A total of 1,509 genes of P. proliferus homologous to traditional VFs, including surface antigens (SAGs), secreted proteins (SPs), ATP-Binding Cassette (ABC) Transporters, actin-related proteins (ARPs), aminopeptidases (APases), glycoproteins (GPs), cysteine proteases (CPs), and heat shock proteins (HSPs), were identified. Meanwhile, homology analysis identified 6279 DEHGs among the five species, of which there were 48 DEHGs being mutually differentialexpressed among the four pairs of comparisons, such as MRP, Tuba 3, PI3K, WASF2, ADK, Nop56, DNAH1, PFK-2/FBPase2, Ppp1r7, SSP7. Furthermore, the overlap between the predicted VFs and DEHGs showed 97 genes of the predicted VFs that simultaneously belonged to DEHGs. Strikingly, of these 97 genes, only 26, including Chymotrypsin, Leucine APases, Cathepsin L, HSP 70, and so on, were higher expressed in P. proliferus while all the remaining were lower expressed than in the four other species. Conclusion: This work provides a fundamental context for further studies of the pathogenicity of P. proliferus. Most of the predicted VFs which simultaneously belonged to DEHGs were lower expressed in P. proliferus.

Background: Oral squamous cell carcinoma (OSCC) has been the sixth most common cancer worldwide. Emerging studies showed long non-coding RNAs to play a key role in human cancers. However, the molecular mechanisms underlying the initiation and progression of OSCC remained to be further explored. Objective: The present study aimed to identify differentially expressed lncRNAs and mRNAs in OSCC. Methods: GSE30784 was analyzed to identify differentially expressed lncRNAs and mRNAs in OSCC. Protein-protein interaction network and co-expression network analyses were performed to reveal the potential roles of OSCC related mRNAs and lncRNAs. Results: In the present study, we identified 21 up-regulated lncRNAs and 54 down-regulated lncRNAs in OSCC progression. Next, we constructed a lncRNA related co-expression network in OSCC, which included 692 mRNAs and 2193 edges. Bioinformatics analysis showed that lncRNAs were widely co-expressed with regulating type I interferon signaling pathway, extracellular matrix organization, collagen catabolic process, immune response, ECM-receptor interaction, Focal adhesion, and PI3K-Akt signaling pathway. A key network, including lncRNA C5orf66-AS1, C21orf15, LOC100506098, PCBP1-AS1, LOC284825, OR7E14P, HCG22, and FLG-AS1, was found to be involved in the regulation of immune response to tumor cell, Golgi calcium ion transport, negative regulation of vitamin D receptor signaling pathway, and glycerol- 3-phosphate catabolic process. Moreover, we found higher expressions of CYP4F29P, PCBP1- AS1, HCG22, and C5orf66-AS1, which were associated with shorter overall survival time in OSCC samples. Conclusions: Our analysis can provide novel insights to explore the potential mechanisms underlying OSCC progression.

Background: Osteosarcoma (OS) is one of the most common primary malignant bone tumors in children and adolescents. OS metastasis has been a challenge in the treatment of OS. The present study screened progression related genes in OS by analyzing a public dataset GSE42352, and identified 691 up-regulated and 945 down-regulated genes in advanced stage OS compared to early-stage OS samples. Methods: Protein-protein interaction (PPI) networks were further employed to reveal the interaction among these genes. Bioinformatics analysis showed that progression related differently expressed genes (DEGs) were significantly associated with the regulation of cell proliferation and metabolisms. Results: This study revealed that progression related DEGs were dysregulated in metastatic OS compared to non-metastatic OS samples. Further analysis showed CSF1R, CASP1, CD163, AP1B1, LAPTM5, PEX19, SLA, STAB1, YWHAH, PLCB2, and GPR84 were associated with the metastasis-free survival time in patients with OS. Conclusion: These findings provided novel information for us to understand the mechanisms underlying the progression and metastasis of OS.

Objectives: Graft-versus-host disease (GVHD) is the most common complication after hematopoietic stem cell transplantation (HSCT) and remains to be a major cause of mortality. Activation of toll-like receptor 4 (TLR-4) by lipopolysaccharide induces the NF-ΚB signaling pathway to release critical proinflammatory cytokines and increases the recipient response to GVHD. In order to clarify the role of TLR-4 in the occurrence of acute GVHD after HSCT, we collected 208 samples from HSCT recipients and their human lecucyte antigen identical donors to test the hypothesis that TLR-4polymorphism in the recipients or donors influence the risk of acute GVHD in allogeneic HSCT recipients. Methods: TLR-4 Asp299Gly and Thr399Ile polymorphisms of each sample were examined by using DNA sequencing and polymerase chain reaction-restriction fragment length polymorphism methods. Results: No homozygous or heterozygous variant alleles of the Asp299Gly or Thr339Ile polymorphism were detected in any samples in our study. Our results demonstrate that TLR-4 Asp299Gly and Thr399Ile polymorphisms might be very rare in the Chinese population (Eastern China and Taiwan region). Conclusion: The results of this study cannot confirm the role of TLR-4 mutations in the pathogenesis of GVHD in humans, yet we reach a definite conclusion by a TLR-4 knockout murine GVHD model in our ongoing project.

Background: Localized radiation therapy is the first-line option for the treatment of nonmetastatic prostate cancer (PCa). Previous studies revealed that long non-coding RNAs (lncRNAs) had crucial roles in disease progression. However, the mechanisms of lncRNAs underlying prostate cancerrelated fatigue remained largely unclear. Objective: The present study aimed to uncover the key genes related to PCa related fatigue during localized radiation therapy by constructing mRNA and lncRNA regulatory networks. Methods: We analyzed GSE30174, which included 10 control samples and 40 PCa related fatigue samples, to identify differently expressed lncRNAs and mRNAs in PCa related fatigue. A proteinprotein interaction network was constructed to reveal the interactions among mRNAs. Co-expression network analysis was applied to identify the key lncRNAs and reveal the functions of these lncRNAs in PCa related fatigue. Results and Discussion: This research found 1271 dysregulated mRNAs and 205 dysregulated lncRNAs in PCa related fatigue using GSE30174. Bioinformatics analysis showed that PCa related fatigue with mRNAs and lncRNAs were associated with inflammatory response and immune response related biological processes. Furthermore, we constructed a PPI network and lncRNA co-expression network related to fatigue in PCa. Of note, we observed that the dysregulated lncRNAs and mRNAs, such as SEC61A2, ADCY6, LPAR5, COL7A1, ALB, COL1A1, SNHG1, LINC01215, LINC00926, GNG4, LMO7, and COL4A6, in PCa related fatigue could predict the outcome of PCa patients. Conclusions: This research could provide novel mechanisms underlying fatigue and identify new biomarkers for the prognosis of PCa.

Background: Bioluminescence is a unique and significant phenomenon in nature. Bioluminescence is important for the lifecycle of some organisms and is valuable in biomedical research, including for gene expression analysis and bioluminescence imaging technology. In recent years, researchers have identified a number of methods for predicting bioluminescent proteins (BLPs), which have increased in accuracy, but could be further improved. Methods: In this study, a new bioluminescent proteins prediction method, based on a voting algorithm, is proposed. Four methods of feature extraction based on the amino acid sequence were used. 314 dimensional features in total were extracted from amino acid composition, physicochemical properties and k-spacer amino acid pair composition. In order to obtain the highest MCC value to establish the optimal prediction model, a voting algorithm was then used to build the model. To create the best performing model, the selection of base classifiers and vote counting rules are discussed. Results and Conclusion: The proposed model achieved 93.4% accuracy, 93.4% sensitivity and 91.7% specificity in the test set, which was better than any other method. A previous prediction of bioluminescent proteins in three lineages was also improved using the model building method, resulting in greatly improved accuracy.

Background: Neuroimaging is an important tool in early detection of Alzheimer’s disease (AD), which is a serious neurodegenerative brain disease among the elderly subjects. Independent component analysis (ICA) is arguably one of the most widely used algorithm for the analysis of brain imaging data, which can be used to extract intrinsic networks of brain from functional magnetic resonance imaging (fMRI). Methods: Witnessed by recent studies, a more flexible model known as restricted Boltzmann machine (RBM) can also be used to extract spatial maps and time courses of intrinsic networks from resting state fMRI, moreover, RBM shows superior temporal features than ICA. Here, we seek to employ RBM to improve the performance of classifying individuals. Experiments are performed on healthy controls and subjects at the early stage of AD, i.e., cognitive normal (CN) and early mild cognitive impairment participants (EMCI), and two types of data, i.e., structural magnetic resonance imaging (sMRI) and fMRI data. Results: (1) By separately employing ICA for sMRI and fMRI, the features extracted from fMRI improve classification accuracy by 7.5% for CN and EMCI; (2) instead of applying ICA to fMRI, using RBM further improves classification accuracy by 7.75% for CN and EMCI; (3) the lesions at the early stage of AD are more likely to occur in the regions around slices 4, 6, 10, 14, 19, 51 and 59 of the whole brain in the longitudinal direction. Conclusion: By using fMRI instead of sMRI and RBM instead of ICA, we can classify CN and EMCI more efficiently.

Background: The advancement of bioinformatics and machine learning has facilitated the diagnosis of cancer and the discovery of omics-based biomarkers. Objective: Our study employed a novel data-driven approach to classifying the normal samples and different types of gastrointestinal cancer samples, to find potential biomarkers for effective diagnosis and prognosis assessment of gastrointestinal cancer patients. Methods: Different feature selection methods were used, and the diagnostic performance of the proposed biosignatures was benchmarked using support vector machine (SVM) and random forest (RF) models. Results: All models showed satisfactory performance in which Multilabel-RF appeared to be the best. The accuracy of the Multilabel-RF based model was 83.12%, with precision, recall, F1, and Hamming- Loss of 79.70%, 68.31%, 0.7357 and 0.1688, respectively. Moreover, proposed biomarker signatures were highly associated with multifaceted hallmarks in cancer. Functional enrichment analysis and impact of the biomarker candidates in the prognosis of the patients were also examined. Conclusion: We successfully introduced a solid workflow based on multi-label learning with High- Throughput Omics for diagnosis of cancer and identification of novel biomarkers. Novel transcriptome biosignatures that may improve the diagnostic accuracy in gastrointestinal cancer are introduced for further validations in various clinical settings.

Background: Detecting DNA-binding proteins (DBPs) based on biological and chemical methods is time-consuming and expensive. Objective: In recent years, the rise of computational biology methods based on Machine Learning (ML) has greatly improved the detection efficiency of DBPs. Methods: In this study, the Multiple Kernel-based Fuzzy SVM Model with Support Vector Data Description (MK-FSVM-SVDD) is proposed to predict DBPs. Firstly, sex features are extracted from the protein sequence. Secondly, multiple kernels are constructed via these sequence features. Then, multiple kernels are integrated by Centered Kernel Alignment-based Multiple Kernel Learning (CKA-MKL). Next, fuzzy membership scores of training samples are calculated with Support Vector Data Description (SVDD). FSVM is trained and employed to detect new DBPs. Results: Our model is evaluated on several benchmark datasets. Compared with other methods, MKFSVM- SVDD achieves best Matthew's Correlation Coefficient (MCC) on PDB186 (0.7250) and PDB2272 (0.5476). Conclusion: We can conclude that MK-FSVM-SVDD is more suitable than common SVM, as the classifier for DNA-binding proteins identification.

Background: Dry Weight (DW) is the lowest weight after dialysis, and patients with lower weight usually have symptoms of hypotension and shock. Several clinical-based approaches have been presented to assess the dry weight of hemodialysis patients. However, these traditional methods all depend on special instruments and professional technicians. Objective: In order to avoid this limitation, we need to find a machine-independent way to assess dry weight, therefore we collected some clinical influencing characteristic data and constructed a Machine Learning-based (ML) model to predict the dry weight of hemodialysis patients. Methods: In this paper, 476 hemodialysis patients' demographic data, anthropometric measurements, and Bioimpedance spectroscopy (BIS) were collected. Among them, these patients' age, sex, Body Mass Index (BMI), Blood Pressure (BP) and Heart Rate (HR) and Years of Dialysis (YD) were closely related to their dry weight. All these relevant data were used to enter the regression equation. Multiple Kernel Support Vector Regression-based on Maximizes the Average Similarity (MKSVRMAS) model was proposed to predict the dry weight of hemodialysis patients. Results: The experimental results show that dry weight is positively correlated with BMI and HR. And age, sex, systolic blood pressure, diastolic blood pressure and hemodialysis time are negatively correlated with dry weight. Moreover, the Root Mean Square Error (RMSE) of our model was 1.3817. Conclusion: Our proposed model could serve as a viable alternative for dry weight estimation of hemodialysis patients, thus providing a new way for clinical practice.

Background: Among all the major Post-translational modification, lipid modifications possess special significance due to their widespread functional importance in eukaryotic cells. There exist multiple types of lipid modifications and Palmitoylation, among them, is one of the broader types of modification, having three different types. The N-Palmitoylation is carried out by attachment of palmitic acid to an N-terminal cysteine. Due to the association of N-Palmitoylation with various biological functions and diseases such as Alzheimer’s and other neurodegenerative diseases, its identification is very important. Objective: The in vitro, ex vivo and in vivo identification of Palmitoylation is laborious, time-taking and costly. There is a dire need for an efficient and accurate computational model to help researchers and biologists identify these sites, in an easy manner. Herein, we propose a novel prediction model for the identification of N-Palmitoylation sites in proteins. Methods: The proposed prediction model is developed by combining the Chou’s Pseudo Amino Acid Composition (PseAAC) with deep neural networks. We used well-known deep neural networks (DNNs) for both the tasks of learning a feature representation of peptide sequences and developing a prediction model to perform classification. Results: Among different DNNs, Gated Recurrent Unit (GRU) based RNN model showed the highest scores in terms of accuracy, and all other computed measures, and outperforms all the previously reported predictors. Conclusion: The proposed GRU based RNN model can help to identify N-Palmitoylation in a very efficient and accurate manner which can help scientists understand the mechanism of this modification in proteins.

Background: Since bacteria are the earliest known organisms, there has been significant interest in their variety and biology, most certainly concerning human health. Recent advances in Metagenomics sequencing (mNGS), a culture-independent sequencing technology, have facilitated an accelerated development in clinical microbiology and our understanding of pathogens. Objective: For the implementation of mNGS in routine clinical practice to become feasible, a practical and scalable strategy for the study of mNGS data is essential. This study presents a robust automated pipeline to analyze clinical metagenomic data for pathogen identification and classification. Methods: The proposed Clin-mNGS pipeline is an integrated, open-source, scalable, reproducible, and user-friendly framework scripted using the Snakemake workflow management software. The implementation avoids the hassle of manual installation and configuration of the multiple commandline tools and dependencies. The approach directly screens pathogens from clinical raw reads and generates consolidated reports for each sample. Results: The pipeline is demonstrated using publicly available data and is tested on a desktop Linux system and a High-performance cluster. The study compares variability in results from different tools and versions. The versions of the tools are made user modifiable. The pipeline results in quality check, filtered reads, host subtraction, assembled contigs, assembly metrics, relative abundances of bacterial species, antimicrobial resistance genes, plasmid finding, and virulence factors identification. The results obtained from the pipeline are evaluated based on sensitivity and positive predictive value. Conclusion: Clin-mNGS is an automated Snakemake pipeline validated for the analysis of microbial clinical metagenomics reads to perform taxonomic classification and antimicrobial resistance prediction.

Background: Cervical cancer is a highly significant cause of mortality in developing countries, and it is one of the most prominent forms of cancer worldwide. Machine learning techniques have been proven more accurate for the identification of cervical cancer as compared to the manual screening methods like Pap smear and Liquid Cytology Based (LCB) tests. Objective: Primarily, these machine-learning techniques use the images of the cervix for cervical cancer risk analysis; in this article, demographic data and medical records of patients are used to identify major causes of cervical cancer. Furthermore, normal classification methods are used as a usual way of classification when the dataset is balanced as this dataset has abundant examples of negative cases as compared to positive cases On the other hand, traditional binary class classifiers are not sufficient to classify the examples of cervical cancer correctly. Methods: We identified the major causes of cervical cancer by employing multiple machine learning feature selection algorithms. After this selection, we trained different machine learning methods including Decision Trees (DTs), Support Vector Machines (SVMs) and Ensemble Learners using all features as well as these important features. Results and Conclusion: AdaBoost is able to classify instances into healthy and unhealthy classes of this unbalanced dataset with 96% accuracy. Based on this model and significant causes of cervical cancer, we aimed to develop a technique for self-risk assessment of cervical cancer, which women can use to know their chances of being infected from cervical cancer after answering some questions about their demographics and medical history.

Background: Protein secondary structure is vital to predicting the tertiary structure, which is essential in deciding protein function and drug designing. Therefore, there is a high requirement of computational methods to predict secondary structure from their primary sequence. Protein primary sequences represented as a linear combination of twenty amino acid characters and contain the contextual information for secondary structure prediction. Objective and Methods: Protein secondary structure predicted from their primary sequences using a deep recurrent neural network. Protein secondary structure depends on local and long-range residues in primary sequences. In the proposed work, the local contextual information of amino acid residues captures with character n-gram. A dense embedding vector represents this local contextual information. Furthermore, the bidirectional long short-term memory (Bi-LSTM) model is used to capture the long-range contexts by extracting the past and future residues information in primary sequences. Results: The proposed deep recurrent architecture is evaluated for its efficacy for datasets, namely ss.txt, RS126, and CASP9. The model shows the Q3 accuracies of 88.45%, 83.48%, and 86.69% for ss.txt, RS126, and CASP9, respectively. The performance of the proposed model is also compared with other state-of-the-art methods available in the literature. Conclusion: After a comparative analysis, it was observed that the proposed model is performing better in comparison to state-of-art methods.

Aims: To assess the error profile in NGS data, generated from high throughput sequencing machines. Background: Short-read sequencing data from Next Generation Sequencing (NGS) are currently being generated by a number of research projects. Depicting the errors produced by NGS platforms and expressing accurate genetic variation from reads are two inter-dependent phases. It has high significance in various analyses, such as genome sequence assembly, SNPs calling, evolutionary studies, and haplotype inference. The systematic and random errors show incidence profile for each of the sequencing platforms i.e. Illumina sequencing, Pacific Biosciences, 454 pyrosequencing, Complete Genomics DNA nanoball sequencing, Ion Torrent sequencing, and Oxford Nanopore sequencing. Advances in NGS deliver galactic data with the addition of errors. Some ratio of these errors may emulate genuine true biological signals i.e., mutation, and may subsequently negate the results. Various independent applications have been proposed to correct the sequencing errors. Systematic analysis of these algorithms shows that state-of-the-art models are missing. Objective: In this paper, an effcient error estimation computational model called ESREEM is proposed to assess the error rates in NGS data. Methods: The proposed model prospects the analysis that there exists a true linear regression association between the number of reads containing errors and the number of reads sequenced. The model is based on a probabilistic error model integrated with the Hidden Markov Model (HMM). Results: The proposed model is evaluated on several benchmark datasets and the results obtained are compared with state-of-the-art algorithms. Conclusion: Experimental results analyses show that the proposed model efficiently estimates errors and runs in less time as compared to others.