Current Bioinformatics - Volume 17, Issue 8, 2022

Background: Rheumatoid Arthritis (RA) is more common in those who have specific genetic types of Human Leukocyte Antigen (HLA). One of the most important genetic risk factors for RA lies in the HLA-DRB1 locus. Objective: This review aimed to determine which HLA-DRB1 alleles were associated with the risk of RA per allele and phenotype group. Methods: Statistical analyses were performed using RevMan version 5.4.1. Results: The meta-analysis included nine articles that involved 3004 RA patients and 2384 healthy controls. In the allele group, the frequencies of three HLA-DRB1 alleles, HLA-DRB1* 10 (OR = 1.88, 95%CI = 1.25–2.83, p = 0.002), HLA-DRB1* 04 (OR = 2.38, 95%CI = 1.73–3.29, p < 0.00001), and HLA-DRB1* 01 (OR = 1.32, 95%CI = 1.08–1.61, p = 0.006), were considerably higher in RA patients than in controls, and these alleles potentially increased the chance of disease development. Five HLADRB1 alleles (* 03, * 07, * 11, * 13, and * 14), were more prevalent in healthy people than in RA patients and may therefore offer protection against disease onset. Only the DRB1* 04 subtypes, DRB1* 0401 (OR = 1.37, 95 percent CI = 1.05–1.79, p = 0.02) and DRB1* 0404 (OR = 1.73, 95% CI = 1.19–12.53, p = 0.004), showed a significant association with the risk of RA in our pooled effect. Conclusion: Our findings demonstrated a significant relationship between HLA-DRB1 and the risk of RA in various ethnic groups.

Background: Hepatocellular Carcinoma (HCC) is associated with high mortality rates and requires the identification of new therapeutic targets. We sought to develop a nomogram for reliably predicting HCC prognosis. Methods: Gene expression was analyzed in R software, while the hub genes were defined as overlapping candidates across five datasets. A prognostic nomogram was constructed using multivariate Cox analysis and evaluated by receiver operating characteristic curve and concordance index analysis. The fractions of tumor microenvironment cells were determined by using xCell. Hypoxia scores were calculated by single-sample gene set enrichment analysis. Statistically, significance and correlation analyses were processed in R. Results: Tow hub genes were identified, and a prognostic nomogram was established and evaluated in the internal validation dataset (Area Under the Curve [AUC] 0.72, 95% Confidence Interval [CI] 0.63- 0.81) and external cohorts (AUC 0.70, 95% CI 0.55-0.85). The risk scores of the prognostic model were positively and negatively correlated with fractions of the T helper 2 (Th2) cells (R = 0.39, p <0.001) and the hematopoietic stem cells (R = -0.27, p <0.001) and Endothelial Cells (ECs; R = -0.24, p <0.001), respectively. Angiogenesis was more active in the high-risk group, accompanied by increased proliferation of ECs. Furthermore, the significance of Hypoxia-Inducible Factor 1-Alpha (HIF1A) gene-related hypoxia in predicting HCC prognosis was demonstrated. Conclusion: A robust prognostic nomogram for predicting the prognosis of patients with HCC was developed. The results suggested that Th2 cells, VEGF-related angiogenesis and HIF1A-related hypoxia may be promising therapeutic targets for prolonging the overall survival of HCC patients.

Background: The design of novel drugs is vital to combat fatal diseases such as Alzheimer’s. With quantum advances in computational methods, artificial intelligence (AI) techniques have been widely utilized in drug discovery. Since drug design is a protracted and resource-intensive process, extensive research is necessary for building predictive in-silico models to discover new medications for Alzheimer’s. A thorough analysis of models is, therefore, required to expedite the discovery of new drugs. Objective: In this study, the performance of machine learning (ML) and deep learning (DL) models for predicting the bioactivity of compounds for Alzheimer’s inhibition is assessed. Additionally, an interaction network is constructed to visualize the clustered bioactivity networks. Methods: The dataset was initially prepared from a public repository of bioactive compounds and was curated. Exploratory data analysis was performed to get insights into the gathered data. A bioactivity interaction network was then constructed to detect communities and compute the network metrics. Next, ML and DL models were built, and their hyperparameters were tuned to improve model performance. Finally, the metrics of all the models were compared to identify the best-performing model for bioactivity prediction. Results: The bioactivity network revealed the formation of three communities. The ML models were ranked based on lower error scores, and the best five models were hybridized to create a blended regressor. Subsequently, two DL models, namely a deep neural network (DNN) and long short-term memory with recurrent neural network architecture (LSTM-RNN), were built. The analysis revealed that the LSTM-RNN outperformed all the models analysed in this study. Conclusion: In summary, this study illustrates a bioactivity network and proposes a DL technique to build robust models for in-silico prediction of drug bioactivity against Alzheimer's.

Background: Proteases regulate cell proliferation, cell growth, biological processes, and overall homeostasis. Several proteases are extensively annotated and well-characterized in pathogenic organisms such as bacteria, parasites, and microbial species as anti-bacterial, anti-parasitic and antimicrobial. Several of these proteins are being explored as viable targets for various drug discovery researches in various microbial diseases. Despite multiple studies on pathogen proteases, comprehensive information on pathogen proteases is scattered or redundant, if available. Methods: We have developed a comprehensive and integrative protease database resource, Prot- PathDB, for 23 pathogen species distributed among five taxa, Amoebozoa, Apicomplexa, Heterolob osea, Kinetoplastida and Fungi. ProtPathDB collects and organizes sequences, class division, signal peptides, localization, post-translational modifications, three-dimensional structure and related structural information regarding binding sites, and binding scores of annotated proteases. Results: The ProtPathDB is publicly available at http://bioinfo.icgeb.res.in/ProtPathDB. Conclusion: We believe that the database will be a one-stop resource for integrative and comparative analysis of pathogen proteases to better understand the functions of the microbial proteases and help drug discovery efforts targeting proteases.

Background: Long noncoding RNAs (LncRNAs) represent a large category of functional RNA molecules that play a significant role in human cancers. lncRNAs can be gene modulators to affect the biological process of multiple cancers. Methods: Here, we developed a computational framework that uses the lncRNA-mRNA network and mutations in individual genes of 9 cancers from TCGA to prioritize cancer lncRNA modulators. Our method screened risky cancer lncRNA regulators based on integrated multiple lncRNA functional networks and 3 calculation methods. Results: Validation analyses revealed that our method was more effective than prioritization based on a single lncRNA network. This method showed high predictive performance, and the highest ROC score was 0.836 in breast cancer. It’s worth noting that we found that 5 lncRNAs scores were abnormally high, and these lncRNAs appeared in 9 cancers. By consulting the literature, these 5 lncRNAs were experimentally supported lncRNAs. Analyses of prioritizing lncRNAs reveal that these lncRNAs are enriched in various cancer-related biological processes and pathways. Conclusion: Together, these results demonstrated the ability of this method to identify candidate lncRNA molecules and improved insights into the pathogenesis of cancer.

Background: Wujiayizhi granule (WJYZG) is a kind of traditional Chinese medicine, which is used for treating Alzheimer's disease (AD). Although the clinical effect of WJYZG for AD is obvious, its underlying mechanism is still obscure. Objective: Explore the mechanism of WJYZG in the treatment of AD by using bioinformatics methods. Methods: Traditional Chinese Medicine Systems Pharmacology Database and Analysis Platform (TCMSP), Traditional Chinese Medicine Integrated Database (TCMID) and Encyclopedia Database of Chinese Medicine (ETCM) were used to search the ingredients and targets of WJYZG. DisGeNET, Drugbank, Online Mendelian Inheritance in Man (OMIM), and Terapeutic Target Database (TTD) were used to retrieve the targets of AD. The Cytoscape3.6.1 software was used to construct the interaction network of herbs-ingredients-targets. The Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses were performed to explore the treatment mechanism of WJYZG on AD. Molecular docking was used to validate the interactions between the ingredients and targets. Results: One hundred and thirty-three ingredients were identified from WJYZG. According to the herbingredient- targets network, quercetin, kaempferol, luteolin, anhydroicaritin, and 8-prenyl-flavone were screened out as the key ingredients, which can interact with the core targets encompassing INS, IL6, TNF, IL1B, CASP3, PTGS2, VEGFA, and PPARG. The enrichment analysis indicates that the treatment of AD by WJYZG was through inhibiting inflammation and neurocyte apoptosis, regulating the calcium ion signaling pathway and adjusting INS levels. Conclusion: The underlying mechanisms of WJYZG in the treatment of AD were theoretically illustrated. We hope these results will enlighten the researches on AD.

Background: Oral cavity cancer (OCC) is one of the most common carcinoma diseases. Recent genome-wide association studies (GWAS) have reported numerous genetic variants associated with OCC susceptibility. However, the regulatory mechanisms of these genetic variants underlying OCC remain largely unclear. Objective: This study aimed to identify OCC-related genetics risk genes contributing to the prognosis of OCC. Methods: By combining GWAS summary statistics (N = 4,151) with expression quantitative trait loci (eQTL) across 49 different tissues from the GTEx database, we performed an integrative genomics analysis to uncover novel risk genes associated with OCC. By leveraging various computational methods based on multi-omics data, we prioritized some of these risk genes as promising candidate genes for drug repurposing in OCC. Results: Using two independent computational algorithms, we found that 14 risk genes whose geneticsmodulated expressions showed a notable association with OCC. Among them, nine genes were newly identified, such as IRF4 (P = 2.5×10^-9 and P = 1.06×10^-4), TNS3 (P = 1.44×10^-6 and P = 4.45×10^-3), ZFP90 (P = 2.37×10^-6 and P = 2.93×10^-4), and DRD2 (P = 2.0×10^-5 and P = 6.12×10^-3), by using MAGMA and S-MultiXcan methods. These 14 genes were significantly overrepresented in several cancer- related terms (FDR < 0.05), and 10 of 14 genes were enriched in 10 potential druggable gene categories. Based on differential gene expression analysis, the majority of these genes (71.43%) showed remarkable differential expressions between OCC patients and paracancerous controls. By integration of multi-omics-based evidence from genetics, eQTL, and gene expression, we identified that the novel risk gene of IRF4 exhibited the highest ranked risk score for OCC (score = 4). Survival analysis showed that dysregulation of IRF4 expression was significantly associated with cancer patients’ outcomes (P = 8.1×10^-5). Conclusion: Based on multiple omics data, we constructed a computational framework to pinpoint risk genes for OCC, and we prioritized 14 risk genes associated with OCC. There were nine novel risk genes, including the IRF4 gene, which is significantly associated with the prognosis of OCC. These identified genes provide a drug repurposing resource to develop therapeutic drugs for treating patients, thereby contributing to the personalized prognostic management of OCC patients.

Background: The odors we perceive are primarily the result of a mixture of odorants. There can be one or multiple odors associated with an odorant. Several studies have attempted to link odorant physicochemical properties to specific olfactory perception; however, no universal rule that can determine how and to what extent molecular properties affect odor perception exists. Objective: This study aims to identify important and common features of odorants with seven basic odors (floral, fruity, minty, nutty, pungent, sweet, woody) to comprehend the complex topic of odors better. Methods: We adopted an in-silico approach to study key and common odorants features with seven fundamental odors (floral, fruity, minty, nutty, pungent, sweet, and woody). A dataset of 1136 odorants having one of the odors was built and studied. Results: A set of nineteen structural features has been proposed to identify seven fundamental odors rapidly. The findings also indicated associations between odors, and specific molecular features associated with each group of odorants and shared spatial distribution of odor features. Conclusion: This study revealed olfactory associations, unique chemical properties linked with each set of odorants, and a common spatial distribution of odor features for considered odors.

Background: Connecting genes to phenotypes is still a great challenge in genetics. Research related to gene-phenotype associations has made remarkable progress recently due to high-throughput sequencing technology and genome-wide association study (GWAS). However, these genes, which are considered to be significantly associated with a target phenotype according to traditional GWAS, are less precise or subject to greater confounding. Objective: The present study is an attempt to prioritize functional genes for complex phenotypes employing protein-protein interaction (PPI) network-based systems genetics methods on available GWAS results. Methods: In this paper, we calculated the functional gene enrichment ratios of the trait ontology of A. thaliana for three common systems genetics methods (i.e. GeneRank, K-shell and HotNet2). Then, comparison of gene enrichment ratios obtained by PPI network-based methods was performed. Finally, a hybrid model was proposed, integrating GeneRank, comprehensive score algorithm and HotNet diffusion- oriented subnetworks (HotNet2) to prioritize functional genes. Results: These PPI network-based systems genetics methods were indeed useful for prioritizing 775henoltype-associated genes. And functional gene enrichment ratios calculated from the top 20% of GeneRank-identified genes were higher than these ratios of K-shell and these ratios of HotNet2 for most phenotypes. However, the hybrid model can improve the efficiency of functional gene enrichment for A. thaliana (up to 40%). Conclusion: The present study provides a hybrid method integrating GeneRank, comprehensive score algorithm and HotNet2 to prioritize functional genes. The method will contribute to functional genomics in plants. The source data and codes are freely available at http://47.242.161.60/Plant/.