Abstract
Background
Inclusion Body Myositis is an acquired muscle disease. Its pathogenesis is unclear due to the co-existence of inflammation, muscle degeneration and mitochondrial dysfunction. We aimed to provide a more advanced understanding of the disease by combining multi-omics analysis with prior knowledge. We applied molecular subnetwork identification to find highly interconnected subnetworks with a high degree of change in Inclusion Body Myositis. These could be used as hypotheses for potential pathomechanisms and biomarkers that are implicated in this disease.
Results
Our multi-omics analysis resulted in five subnetworks that exhibit changes in multiple omics layers. These subnetworks are related to antigen processing and presentation, chemokine-mediated signaling, immune response-signal transduction, rRNA processing, and mRNA splicing. An interesting finding is that the antigen processing and presentation subnetwork links the underexpressed miR-16-5p to overexpressed HLA genes by negative expression correlation. In addition, the rRNA processing subnetwork contains the RPS18 gene, which is not differentially expressed, but has significant variant association. The RPS18 gene could potentially play a role in the underexpression of the genes involved in 18 S ribosomal RNA processing, which it is highly connected to.
Conclusions
Our analysis highlights the importance of interrogating multiple omics to enhance knowledge discovery in rare diseases. We report five subnetworks that can provide additional insights into the molecular pathogenesis of Inclusion Body Myositis. Our analytical workflow can be reused as a method to study disease mechanisms involved in other diseases when multiple omics datasets are available.
Supplementary Information
The online version contains supplementary material available at 10.1186/s13023-024-03526-x.
Keywords: Inclusion body myositis, Multi-omics, Transcriptomics, Genomics, Network analysis, Active subnetwork identification, Rare diseases, Data integration, Multiplex network
Background
Inclusion body Myositis (IBM) is a rare, acquired muscle disease with a prevalence ranging from 24.8 to 45.6 per million people [1], but the most common muscle disease with onset after age 50 [2]. The molecular pathogenesis of IBM has consistently been of high interest due to the unexplained combination of inflammatory changes, degenerative features, and mitochondrial abnormalities in the muscle tissue [3, 4]. One hypothesis suggests that autoimmunity drives protein aggregation, resulting in high interferon-gamma and cytotoxic T-cell responses [5]. Endomysial infiltration of CD8 + T cells in IBM muscles and the reported clonal expansion of these T cells in the blood and muscles of IBM patients suggest the presence of unknown antigens [6]. Currently, the links between these antigens, protein aggregate pathology, and the immune response are poorly understood. Identifying these links could contribute to understanding the disease pathomechanisms and thereby to the development of more effective diagnosis and treatment.
The integrated analysis of multiple types of omics datasets (multi-omics analysis) may provide new insights into potential disease causing mechanisms as well as knowledge about their interplay. A single omics layer often provides information about a single aspect of one type of molecule. For instance, in processed transcriptomics datasets, the abundance of the mRNA is usually the only information provided. Integrating this data with other omics, like genomics and microRNA (miRNA) transcriptomics, can provide a better picture of the molecular state of cells in the disease. In addition, a multi-omics approach can increase the statistical power of analyses, even when the number of available patient samples is limited due to the rare occurrence of the disease [7]. Finally, multi-omics approaches can give more insight into the flow of information in the disease, for example, from genetic factors to their consequences [8].
In biological research, there is an abundance of knowledge available from previous experiments and research, such as protein-protein interactions and pathway information. This prior knowledge can be used in research, and benefit the study of rare diseases by allowing more and diverse information to be used in the analysis, despite the limited number of patient samples. To fully exploit prior knowledge and multi-omics data analysis, we combined these two to increase the amount of information available. This has several benefits [9] (i) focusing the analysis on the results that are more likely to be biologically relevant, (ii) deprioritizing spurious results arising from noise instead of biological signals, since they are less likely to be associated with prior knowledge, and (iii) providing extra knowledge and data together with the results, which supports the formulation of hypotheses after analysis.
In this work, we performed an integrative multi-omics data analysis with prior knowledge to investigate mechanisms that are disrupted in IBM. We created a large-scale network combining different types of interactions involving genes/proteins and miRNAs. Using an active subnetwork identification algorithm, we identified several subnetworks that were highly relevant for IBM and reflected processes that are already known to be affected in IBM, but also some novel ones.
Methods
Workflow
We implemented a workflow that consists of multiple steps, namely differential expression testing, burden testing, network construction, active subnetwork identification and functional profiling. An overview of the workflow is shown in Fig. 1. To make the workflow more Findable, Accessible, Interoperable and Reusable (FAIR) [10], we made it available on WorkflowHub [11]. The workflow is developed in Common Workflow Language (CWL) [12], which ensures the scripts always run the same way when reused. It also allows metadata to be embedded in the inputs, outputs and steps of the workflow for the purpose of findability and reusability. Finally, we made a Docker container [13], and attached it to the workflow to ensure our computational environment is reproduced upon reuse.
Fig. 1.
The overview of the complete workflow. The IBM data is combined with knowledge from databases in order to construct a network. Active subnetwork identification is applied on this network to find active subnetworks. These subnetworks are subsequently annotated with Gene Ontology annotation
RNA-Seq differential expression analysis
We obtained gene expression count datasets (both mRNA and miRNA) from the Gene Expression Omnibus database (GEO) using the GSE151758 accession code, which have samples for IBM patients and amputee controls. This dataset was generated by Johari et al. [14] using short-read polyA + RNA sequencing from muscle biopsy tissues. To ensure that we have an integrated network in which the miRNA and mRNA results reflect the same biological changes, we used only the eighteen IBM and nine control samples which had both mRNA and miRNA data available.
We repeated the differential gene expression analysis (with the original script) that was performed by Johari et al. [14] in R (version 4.0.5) [15, 16], using Bioconductor (version 1.30.10) [17] and DESeq2 (version 1.30.1) [18]. With DESeq2, the raw data was transformed to be approximately homoskedastic and normalized for factor and library size. The data was then fitted to a Negative Binomial Generalized Linear Model and tested for differential expression between the IBM and amputee cohorts using the Wald test (additional file 1 and 2).
Exome sequencing variant burden test
We expanded upon our previously published Finnish IBM cohort [19], bringing the total to 81 Finnish IBM patients. For the 51 additional individuals, we produced exome sequencing data as described previously [19]. This dataset includes the 18 Finnish IBM patients for whom we also have mRNA and miRNA data available. We used Finnish controls (n = 99) from 1000 Genomes project and downloaded the exome sequencing datasets for the same. We then generated genotypes in VCF format for both cases and controls. To increase the statistical power of the analysis, we performed a rare variant burden test using the “RVTESTS” software (version 2.1.0) [20, 21], suitable for testing rare variants with different directions of effects. We used RVTESTS with the following parameters: Burden = CMC (Combined Multivariate and Collapsing), Kernel = SKAT (Sequence Kernel Association Test), and Variable threshold model by permutation (price). We used the resulting p-values from the SKAT test in subsequent analysis.
IBM multi-omics multiplex network construction
We created a multiplex multi-omics IBM network comprising two layers (Fig. 2). The first layer is composed of protein-protein interactions that have an experimental evidence score of at least 0.200 (to filter out low-confidence interactions) in STRING [22] (version 11.0) and miRNA-mRNA target pairs from miRTarBase [23] (version 8.0). The second layer comprises mRNA-mRNA and miRNA-mRNA pairs, with their biweight midcorrelation calculated from the normalized transcriptomics data using the equation defined in [24]. In our calculation, only samples that had data in both transcriptomics datasets, and only genes with expression in more than half of the samples were used. In order to capture the correlations (and directions of correlations) with the most biologically informative information, the correlations were transformed to a binary form by setting a threshold of > 0.7 for mRNA-mRNA pairs and a more relaxed threshold of <-0.5 for miRNA-mRNA pairs (lower mean correlation). In the resulting multiplex multi-omics network, a node represents either a miRNA or a gene. Note that the gene node corresponds to both the mRNA and the protein. Regarding the miRNA nodes, we assigned a p-value on each node based on their differential gene expression. Similarly, for each gene node we assigned a combined p-value using Fisher’s method based on the differential gene expression and the variant burden of each particular gene. Note that this p-value is only valid for prioritization and not for statistical inference, since different null hypotheses are combined.
Fig. 2.
The workflow for the network creation. Step 1: Interaction data is downloaded from STRING DB and miRTarBase. Expression correlation is calculated from the IBM expression datasets. These are mapped and combined into a database layer, and a correlation layer. Step 2: Differential gene expression testing is applied to the mRNA and miRNA data. Burden testing is applied to the exome sequencing data. In the case of mRNAs, the two p-values are combined into one. Step 3: The node values are overlaid on top of the network created from the edges
Finally, to connect everything in the network, we used BridgeDB [25] and miRBaseConverter [26] to map the identifiers of STRING and miRTarBase to the identifiers of the transcriptome and exome datasets. Specifically, we mapped NCBI gene identifiers to Ensembl identifiers, and mapped miRNA names to miRBase accessions. We removed isolated nodes that were not connected to the largest connected component of the network, because we are using an iterative analysis method that traverses the network. The resulting network consists of 17,405 nodes and 3,353,996 edges of different types (Table 1).
Table 1.
Summary statistics for the multiplex network
| Statistic | Count |
|---|---|
| Number of nodes | 17,405 |
| Number of gene nodes | 15,574 |
| Number of miRNA nodes | 1,831 |
| Number of genes with variant burden (padj < 0.05) | 209 |
| Number of mRNA differentially expressed (padj < 0.05) | 8215 |
| Number of miRNA differentially expressed (padj < 0.05) | 628 |
| Number of edges | 3,353,996 |
| Number of STRING edges | 300,536 |
| Number of miRTarBase edges | 153,694 |
| Number of mRNA-mRNA correlation edges | 2,157,070 |
| Number of miRNA-mRNA correlation edges | 742,696 |
Active subnetwork identification
We used MOGAMUN [27] to identify highly interconnected subnetworks that have a high degree of biological significance (active subnetworks) using default parameters. Thirty parallel runs were performed with 500 generations of optimization each. Since subnetworks are allowed to overlap, they were merged during post processing in MOGAMUN. In this process, the maximum number of nodes per subnetwork was increased to 200, and the Jaccard Index threshold was lowered to 0.2, in order to obtain subnetworks that are highly distinctive. We performed functional profiling of each resulting subnetwork with Gene Ontology Biological Processes terms using the g: GOSt function in the gProfiler2 R package (version 2.0) [28]. We also calculated the correlation of each subnetwork’s first principal component with the estimation of several cell types as estimated by Johari et al. [14, 29, 30].
Results
In order to elucidate IBM disease mechanisms from multi-omics data, we applied active subnetwork identification [27] on our IBM network. This resulted in five distinct subnetworks which we labeled according to their top GO terms: “Antigen processing and presentation”, “Chemokine-mediated signaling”, “Immune response – signal transduction”, “rRNA processing”, and “mRNA splicing”, respectively (Fig. 3). These subnetworks are optimized based on both the density of interactions and the scores (p-values) of the nodes. All p-values, correlations and fold changes in these subnetworks are available in additional file 3 and 4.
Fig. 3.
The five subnetworks that were identified by the active subnetwork identification algorithm together with their top-scoring GO annotations. Node shapes indicate the type of each entity. The gene and miRNA expression fold change is shown using color (red indicates upregulation; dark red indicates a fold change > 5.5; blue indicates downregulation; white indicates no change). Variant burden significance is shown with a solid line around a node. The different types of an edge are depicted with the line type
The “Antigen processing and presentation” subnetwork (Fig. 3a) contains 20 nodes, of which 19 are genes, and one is a miRNA. These 20 nodes are all significantly differentially expressed (padj < 0.05) and two of those (HLA-C and BAG6) also have a significant variant burden (padj = 0.013 and padj = 0.008, respectively). Six nodes are overexpressed HLA genes (HLA-A, HLA-B, HLA-C, HLA-E, HLA-F, and HLA-G). In previous studies, their overexpression was consistent in IBM and other inflammatory myopathies [14, 31]. This subnetwork also contains several overexpressed CD genes, namely CD2, CD5, CD48, CD79A, CD8A, and CD8B, which are markers of Leukocytes [32]. LCK was another overexpressed gene in this subnetwork, previously proposed as an apoptosis regulator involved in IBM [14]. Besides protein-coding genes, this subnetwork contains the miRNA miR-16-5p, linked to the HLA-A, HLA-B, and HLA-C genes, with negative correlation coefficients ranging from − 0.51 to -0.58. Finally, this subnetwork is strongly correlated to the proportion estimation of CD8 T cells, CD4 memory activated T cells, regulatory T cells (additional file 5).
The “Chemokine-mediated signaling” subnetwork (Fig. 3b) contains 29 genes and three miRNAs. Most of the genes involved in this subnetwork are overexpressed (padj < 0.05). These genes mainly include various cytokines, specifically chemokine ligands and receptors such as CCL5, CCR4, CXCL10, and XCL1. Many of these chemokines are involved in calcium signaling, which has been hypothesized by Johari et al. to play a role in IBM [14].
The “Immune response - signal transduction” subnetwork (Fig. 3c) shares several genes with the previous two subnetworks, though the overlap is below the merging threshold. It shares LCK, CD2, CD8A, UBC, and PTPRC with the “antigen processing and presentation” subnetwork and CD4, UBC, and SYK with the “chemokine-mediated signaling” subnetwork, connecting to these two subnetworks. Only the CD247, ZAP70, PIK3R1, VAV1, STAT1, CD28, FYN, LCP2 and RACK1 genes are unique to this subnetwork. Genes such as LCK, FYN, ZAP70 and VAV1 are part of T cell receptor signaling [33], which is triggered by binding of the T cell receptor to antigen presenting MHC complexes [34].
The “rRNA processing” subnetwork (Fig. 3d) contains 15 genes. In this subnetwork, the RPS18 (ribosomal protein S18) gene has a significant variant burden (padj = 0.046) but, interestingly, is not significantly differentially expressed (padj = 0.22). The rest of the genes in this subnetwork are downregulated (padj < 0.05) with no significant variant burden. These include genes like WDR43, IMP4, NOP14, RRP9, PDCD11, UTP18, and UTP15, which have a role in the processing of the 18 S ribosomal RNA [35–41].
Finally, the last subnetwork is annotated with “mRNA splicing” (Fig. 3e). Although five of the 38 genes in this subnetwork (ASPN, LRRC17, PPIC, DOCK4, and TLR3) are overexpressed (padj < 0.05), the rest of the genes are underexpressed (padj < 0.05). Many of these genes, such as DDX21, RBM8A, CWC25 and EFTUD2 are involved in mRNA splicing [42–45]. This subnetwork also contains genes with a significant variant burden, namely SNAPC4 and DOCK4.
Discussion
This work presents the identification of multi-omics signatures in IBM that provide insights into potential disease mechanisms that are at play. We identified five subnetworks that represent these signatures. Within these subnetworks, we find several interesting interactions which can be the basis for forming hypotheses in IBM.
Many of our results correspond to earlier findings. For example, the HLA genes in the “Antigen processing and presentation” subnetwork are consistently found to be overexpressed in IBM [14, 31]. LCK was another overexpressed gene in this subnetwork, previously proposed as an apoptosis regulator involved in IBM [14].
There are also novel findings, such as the miR-16-5p. This miRNA was connected with many HLA genes in the “Antigen processing and presentation” subnetwork. Here, the downregulation of miR-16-5p was negatively correlated with the upregulation of the HLA genes, which is coherent with the canonical mechanism of miRNAs suppressing gene expression. Interestingly, the overexpression of miR-16 and miR-15a led to a significantly decreased pro-inflammatory signaling through IL-1β, TNFα, and NF-κB in a study in mice [46]. In addition, another study in mice links the conditional deletion of these microRNAs to proliferation of T regulatory cells (of which the estimated abundance is correlated to several of our subnetworks) and loss of immune tolerance [47]. At the same time, in an epithelial-like cell line, transfection of miR-16 led to upregulation of HLA-G, HLA-A, HLA-B, and HLA-C [48]. We speculate that miR-16 could play a role in regulating MHC class I gene transcription in IBM, which in turn regulates immune system activation. Recently, Lucchini et al. identified dysregulation of hsa-miR-192-5p and hsa-miR-372-3p in serum of IBM patients [49]. Furthermore, miR-16 specifically was implicated in several inflammatory diseases including rheumatoid arthritis, ankylosing Spondylitis and inflammatory bowel disease [50–53]. These findings highlight the importance of studying the role of miRNAs in the context of molecular pathomechanisms of IBM. Interestingly, miR-16 is also connected to SQSTM1, of which the encoded protein is aggregated in IBM and other myopathies [54, 55]. Despite SQSTM1 being a proposed target of miR-16 [49], its expression is not negatively correlated with the miR-16 in our data,perhaps due the influence of other regulatory factors. SQSTM1 was also underexpressed as it was in previous studies with RNA-Seq [55].
The “Chemokine-mediated signaling” subnetwork is interesting because many chemokines are involved in calcium signaling. Disturbed calcium signaling has been proposed as a candidate mechanism in IBM [14, 57]. The prominence of this signature in our results (many strongly dysregulated genes that are interconnected) supports this.
Another gene of interest is RPS18, which encodes a ribosomal protein. Interestingly, some of the surrounding genes are involved in processing the 18 S ribosomal RNA, which is essential for ribosome function [58]. Alterations in RNA metabolism have been implicated in IBM [59]. In addition, in a proteomics study, ribosomal and nuclear proteins were overrepresented in rimmed vacuoles in IBM compared to controls [60]. These findings suggest a role of altered protein synthesis in IBM.
Similarly, the “mRNA splicing” subnetwork, which contains many underexpressed genes, points to a potential dysfunction of the spliceosome in IBM, especially since there is evidence for genes that have altered splicing in IBM [14, 61, 62]. Specifically, the SNAPC4 gene is involved in the transcription of snRNAs that are part of the spliceosome. It has a significant variant burden and thus could play a more causal role in altered splicing. SNAPC4 is also associated with ankylosing spondylitis, an inflammatory disease that affects the spine, and like IBM, has the MHC implicated in its pathogenesis [63].
Finally, our study demonstrates how the application of active subnetwork identification on multi-omics data can connect findings and interactions in different omics, and thereby provide hypotheses about their interplay, An example of this is the link between miR-16-5p and the HLA genes in the antigen processing and presentation subnetwork as shown in our study. In addition, in the rRNA processing subnetwork, we found the RPS18 gene as having a significant variant burden without exhibiting any significant changes in gene expression. We found this gene because it was linked with many downregulated genes that have functions directly related to RPS18. We speculate that variants in RPS18 may affect the expression of the related genes.
Some limitations of our approach are important to note. Although prior knowledge aids in the analysis of omics data in several ways, prior knowledge is also limited by our current understanding of biology and can be biased towards biological concepts that have been studied more. For example, genes that are studied more, such as genes involved in cancer, have more known interactions and are therefore more likely to be overrepresented in network analysis. However, this limitation is mitigated since in our approach we included experimental data in the network. Another limitation primarily affecting multi-omics analysis is the different number of features in each omics. For example, in our analysis, 15,574 nodes represent protein coding genes in our analysis, but only 1831 nodes represent miRNAs. Consequently, relatively few miRNAs were present in the results, which limits the priority of miRNAs in this study.
Our study also has some limitations in terms of data availability. Although for the expression data, a subset was used with a balanced representation in terms of age and sex, the limited sample size still makes our study more sensitive to sampling error caused by individual variation. Further, the accuracy of the cell type proportion through deconvolution estimation is limited since the used reference cell type expression profile (blood) deviates from the muscle disease. Finally, histopathology could not be directly correlated because of subjectivity in sample classification. In order to make also muscle pathology data available as addendum in future studies, streamlining the biopsy procedures and light microscopic analysis processing is recommended.
In future work, the hypotheses surrounding miR-16-5p and the RPS18 gene in IBM could be further studied to increase understanding of the disease and thereby provide opportunities for treatments. Expanding to other diseases, our network based approach can also be applied in diseases with multi-omics data available in order to gain new insights into the interplay between different omics. Finally, the methodology to apply active subnetwork identification on multi-omics datasets Could also be further improved, for example, by exploring normalization and weighting schemes for the multiple omics layers and data sources.
Conclusions
In this work, we present an integrative approach that combines experimental multi-omics data and prior knowledge for elucidating the mechanisms that are implicated in IBM. We identified five subnetworks that combine findings from different omics datasets and interactions. For example, the antigen processing and presentation subnetwork links genes with differential gene expression to genes with significant variant burden and miRNAs. Specifically, the underexpressed miR-16-5p was connected to multiple overexpressed HLA genes by negative expression correlation. This connection could potentially play a role in the regulation of the HLA genes.
Similarly, we found the RPS18 gene having both a variant burden and being connected to many underexpressed genes involved in 18 S ribosomal RNA processing. Mutations in this gene could thus affect the expression of the connected genes and play a role in IBM. Moreover, our analytical workflow which was implemented using the common workflow language can be reused for other case studies.
Electronic supplementary material
Below is the link to the electronic supplementary material.
Additional file 1: Differential gene expression. Table with results of differential gene expression analysis.
Additional file 2: Differential miRNA expression. Table with results of miRNA expression analysis.
Additional file 3: Subnetwork nodes. Table of all nodes in the subnetworks, with their p-values.
Additional file 4: Subnetwork edges. Table of all edges in the subnetwork, with their score or correlation.
Additional file 5: Cell type correlations. Figure with correlations of subnetwork expression to estimated cell type proportions.
Acknowledgements
We would like to thank Manu Jokela and Johanna Palmio for patient consultation and patient material collection. We would also like to thank Elva María Novoa-del-Toro for help with active subnetwork identification. We would like to thank all members of Work Package 13 in the EJP RD consortium for fruitful discussions. We would also like to thank the members of the BioSemantics group for their advice.
Author contributions
DW: Conceptualization, Methodology, Software, Formal analysis, Writing - Original Draft, Visualization. MJ: Conceptualization, Formal analysis, Investigation, Resources, Writing - Review & Editing, Funding acquisition. OO: Conceptualization, Writing - Review & Editing. PACtH: Conceptualization, Writing - Review & Editing. FE: Conceptualization, Writing - Review & Editing. AB: Conceptualization, Writing - Review & Editing. CTE: Conceptualization, Writing - Review & Editing. BU: Writing - Review & Editing. MR: Writing - Review & Editing, Funding acquisition. EM: Conceptualization, Methodology, Writing - Review & Editing, Supervision, Funding acquisition.
Funding
This initiative has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement N°825575 (The European Joint Programme Rare Diseases, EJP RD). This work was carried out on the Dutch national e-infrastructure with the support of SURF Cooperative. MJ received support from the Päivikki and Sakari Sohlberg Foundation, Doctoral School of Health (University of Helsinki), The Biomedicum Helsinki Foundation and the Association Française contre les Myopathies (AFM Téléthon, The French Muscular Dystrophy Association, grant award number: 24438). Access to the Biobank reference panel data for genomic imputation was provided by THL (Finnish Institute for Health and Welfare) under research project BB20202_4.
Data availability
The Workflow and code is available on GitHub (https://github.com/dwijnbergen/IBM_ASI_workflow) and WorkflowHub (10.48546/workflowhub.workflow.681.7). The Docker image is available on https://hub.docker.com/R/jdwijnbergen/multi-omics_asi, and it’s build requirements are available on Zenodo (10.5281/zenodo.10210364). The input data of our workflow is available on Zenodo at 10.5281/zenodo.10411125.
Declarations
Ethics approval and consent to participate
The study was performed in line with the principles of the Declaration of Helsinki. Ethical approval for this study falls under HUS:195/13/03/00/11. Informed consent from the patients was obtained at the time of sample collection.
Consent for publication
Not applicable.
Competing interests
The authors declare that they have no competing interests.
Footnotes
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Daphne Wijnbergen and Mridul Johari contributed equally to this work.
References
- 1.Callan A, Capkun G, Vasanthaprasad V, Freitas R, Needham M. A systematic review and Meta-analysis of Prevalence studies of sporadic inclusion body myositis. J Neuromuscul Dis. 2017;4:127–37. [DOI] [PubMed] [Google Scholar]
- 2.Felice KJ, Whitaker CH, Wu Q, Larose DT, Shen G, Metzger AL, et al. Sensitivity and clinical utility of the anti-cytosolic 5′-nucleotidase 1A (cN1A) antibody test in sporadic inclusion body myositis: report of 40 patients from a single neuromuscular center. Neuromuscul Disord. 2018;28:660–4. [DOI] [PubMed] [Google Scholar]
- 3.Schmidt K, Schmidt J. Inclusion body myositis: advancements in diagnosis, pathomechanisms, and treatment. Curr Opin Rheumatol. 2017;29:632–8. [DOI] [PubMed] [Google Scholar]
- 4.Keller CW, Schmidt J, Lünemann JD. Immune and myodegenerative pathomechanisms in inclusion body myositis. Ann Clin Transl Neurol. 2017;4:422–45. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Snedden AM, Kellett KAB, Lilleker JB, Hooper NM, Chinoy H. The role of protein aggregation in the pathogenesis of inclusion body myositis. Clin Exp Rheumatol. 2022;40:414–24. [DOI] [PubMed] [Google Scholar]
- 6.Greenberg SA. Inclusion body myositis: clinical features and pathogenesis. Nat Rev Rheumatol. 2019;15:257–72. [DOI] [PubMed] [Google Scholar]
- 7.Das S, Mukhopadhyay I. TiMEG: an integrative statistical method for partially missing multi-omics data. Sci Rep. 2021;11:24077. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Hasin Y, Seldin M, Lusis A. Multi-omics approaches to disease. Genome Biol. 2017;18:1–15. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Reshetova P, Smilde AK, van Kampen AH, Westerhuis JA. Use of prior knowledge for the analysis of high-throughput transcriptomics and metabolomics data. BMC Syst Biol. 2014;8:S2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Wilkinson MD, Dumontier M, Aalbersberg IJ, Appleton G, Axton M, Baak A, et al. Comment: the FAIR Guiding principles for scientific data management and stewardship. Sci Data. 2016;3:1–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Inclusion Body Myositis Active Subnetwork Identification Workflow. https://workflowhub.eu/workflows/681. Accessed 13 Dec 2023.
- 12.Crusoe MR, Abeln S, Iosup A, Amstutz P, Chilton J, Tijanić N, et al. Methods included: standardizing computational reuse and portability with the Common Workflow Language. Commun ACM. 2022;65:54–63. [Google Scholar]
- 13.jdwijnbergen/multi-. omics_asi general | Docker Hub. https://hub.docker.com/repository/docker/jdwijnbergen/multi-omics_asi/general. Accessed 13 Dec 2023.
- 14.Johari M, Vihola A, Palmio J, Jokela M, Jonson PH, Sarparanta J, et al. Comprehensive transcriptomic analysis shows disturbed calcium homeostasis and deregulation of T lymphocyte apoptosis in inclusion body myositis. J Neurol. 2022;269:4161–73. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.R Core Team. R: A Language and Environment for Statistical Computing. 2021.
- 16.RStudio Team. RStudio: Integrated Development Environment for R. 2021.
- 17.Gentleman RC, Carey VJ, Bates DM, Bolstad B, Dettling M, Dudoit S, et al. Bioconductor: open software development for computational biology and bioinformatics. Genome Biol. 2004;5:R80. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Love MI, Huber W, Anders S. Moderated estimation of Fold change and dispersion for RNA-seq data with DESeq2. Genome Biol. 2014;15:550. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Johari M, Arumilli M, Palmio J, Savarese M, Tasca G, Mirabella M, et al. Association study reveals novel risk loci for sporadic inclusion body myositis. Eur J Neurol. 2017;24:572–7. [DOI] [PubMed] [Google Scholar]
- 20.Zhan X, Hu Y, Li B, Abecasis GR, Liu DJ. RVTESTS: an efficient and comprehensive tool for rare variant association analysis using sequence data. Bioinformatics. 2016;32:1423–6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Wu MC, Lee S, Cai T, Li Y, Boehnke M, Lin X. Rare-variant association testing for sequencing data with the sequence kernel association test. Am J Hum Genet. 2011;89:82–93. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Szklarczyk D, Gable AL, Lyon D, Junge A, Wyder S, Huerta-Cepas J, et al. STRING v11: protein–protein association networks with increased coverage, supporting functional discovery in genome-wide experimental datasets. Nucleic Acids Res. 2019;47:D607–13. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Huang H-Y, Lin Y-C-D, Li J, Huang K-Y, Shrestha S, Hong H-C, et al. miRTarBase 2020: updates to the experimentally validated microRNA–target interaction database. Nucleic Acids Res. 2020;48:D148–54. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Song L, Langfelder P, Horvath S. Comparison of co-expression measures: mutual information, correlation, and model based indices. BMC Bioinformatics. 2012;13:328. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.van Iersel MP, Pico AR, Kelder T, Gao J, Ho I, Hanspers K, et al. The BridgeDb framework: standardized access to gene, protein and metabolite identifier mapping services. BMC Bioinformatics. 2010;11:5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Xu T, Su N, Liu L, Zhang J, Wang H, Zhang W, et al. miRBaseConverter: an R/Bioconductor package for converting and retrieving miRNA name, accession, sequence and family information in different versions of miRBase. BMC Bioinformatics. 2018;19:514. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Novoa-del-Toro EM, Mezura-Montes E, Vignes M, Térézol M, Magdinier F, Tichit L, et al. A multi-objective genetic algorithm to find active modules in multiplex biological networks. PLOS Comput Biol. 2021;17:e1009263. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Kolberg L, Raudvere U, Kuzmin I, Vilo J, Peterson H. gprofiler2 -- an R package for gene list functional enrichment analysis and namespace conversion toolset g:Profiler. 2020. [DOI] [PMC free article] [PubMed]
- 29.Newman AM, Liu CL, Green MR, Gentles AJ, Feng W, Xu Y, et al. Robust enumeration of cell subsets from tissue expression profiles. Nat Methods. 2015;12:453–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Sturm G, Finotello F, List M, Immunedeconv. An R Package for Unified Access to computational methods for estimating Immune cell fractions from Bulk RNA-Sequencing data. In: Boegel S, editor. Bioinformatics for Cancer Immunotherapy: methods and protocols. New York, NY: Springer US; 2020. pp. 223–32. [DOI] [PubMed] [Google Scholar]
- 31.Nagaraju K, Casciola-Rosen L, Lundberg I, Rawat R, Cutting S, Thapliyal R, et al. Activation of the endoplasmic reticulum stress response in autoimmune myositis: potential role in muscle fiber damage and dysfunction. Arthritis Rheum. 2005;52:1824–35. [DOI] [PubMed] [Google Scholar]
- 32.Bernard A, Boumsell L, Dausset J, Milstein C, Schlossman SF, editors. Leucocyte typing: human leucocyte differentiation antigens detected by monoclonal antibodies. Specification — classification — nomenclature. Berlin, Heidelberg: Springer; 1984. [Google Scholar]
- 33.Hwang J-R, Byeon Y, Kim D, Park S-G. Recent insights of T cell receptor-mediated signaling pathways for T cell activation and development. Exp Mol Med. 2020;52:750–61. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Courtney AH, Lo W-L, Weiss A. TCR Signaling: mechanisms of initiation and propagation. Trends Biochem Sci. 2018;43:108–23. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Bi X, Xu Y, Li T, Li X, Li W, Shao W, et al. RNA targets ribogenesis factor WDR43 to Chromatin for transcription and pluripotency control. Mol Cell. 2019;75:102–e1169. [DOI] [PubMed] [Google Scholar]
- 36.Lee SJ, Baserga SJ. Imp3p and Imp4p, two specific components of the U3 small nucleolar ribonucleoprotein that are essential for Pre-18S rRNA Processing. Mol Cell Biol. 1999;19:5441–52. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Liu PCC, Thiele DJ. Novel stress-responsive genes EMG1 and NOP14 encode conserved, interacting proteins required for 40S Ribosome Biogenesis. Mol Biol Cell. 2001;12:3644–57. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Chen S, Blank MF, Iyer A, Huang B, Wang L, Grummt I, et al. SIRT7-dependent deacetylation of the U3-55k protein controls pre-rRNA processing. Nat Commun. 2016;7:10734. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Sweet T, Khalili K, Sawaya BE, Amini S. Identification of a novel protein from glial cells based on its ability to interact with NF-κB subunits. J Cell Biochem. 2003;90:884–91. [DOI] [PubMed] [Google Scholar]
- 40.Bernstein KA, Gallagher JEG, Mitchell BM, Granneman S, Baserga SJ. The small-subunit processome is a Ribosome Assembly Intermediate. Eukaryot Cell. 2004;3:1619–26. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Prieto J-L, McStay B. Recruitment of factors linking transcription and processing of pre-rRNA to NOR chromatin is UBF-dependent and occurs independent of transcription in human cells. Genes Dev. 2007;21:2041–54. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Gong M, Zhang X, Wang Y, Mao G, Ou Y, Wei C, et al. DDX21 interacts with nuclear AGO2 and regulates the alternative splicing of SMN2. Biosci Biotechnol Biochem. 2021;85:272–9. [DOI] [PubMed] [Google Scholar]
- 43.Chiu Y-F, Liu Y-C, Chiang T-W, Yeh T-C, Tseng C-K, Wu N-Y, et al. Cwc25 is a novel splicing factor required after Prp2 and Yju2 to facilitate the first Catalytic reaction. Mol Cell Biol. 2009;29:5671–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Fabrizio P, Laggerbauer B, Lauber J, Lane WS, Lührmann R. An evolutionarily conserved U5 snRNP-specific protein is a GTP-binding factor closely related to the ribosomal translocase EF-2. EMBO J. 1997;16:4092–106. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45.Zhang X, Yan C, Hang J, Finci LI, Lei J, Shi Y. An atomic structure of the human spliceosome. Cell. 2017;169:918–e92914. [DOI] [PubMed] [Google Scholar]
- 46.Ye E-A, Liu L, Jiang Y, Jan J, Gaddipati S, Suvas S, et al. miR-15a/16 reduces retinal leukostasis through decreased pro-inflammatory signaling. J Neuroinflammation. 2016;13:305. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47.Johansson K, Gagnon JD, Zhou SK, Fassett MS, Schroeder AW, Kageyama R, et al. An essential role for miR-15/16 in Treg suppression and restriction of proliferation. Cell Rep. 2023;42:113298. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 48.Friedrich M, Vaxevanis CK, Biehl K, Mueller A, Seliger B. Targeting the coding sequence: opposing roles in regulating classical and non-classical MHC class I molecules by miR-16 and miR-744. J Immunother Cancer. 2020;8:e000396. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 49.Lucchini M, De Arcangelis V, Santoro M, Morosetti R, Broccolini A, Mirabella M. Serum-circulating microRNAs in sporadic inclusion body myositis. Int J Mol Sci. 2023;24:11139. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 50.Yan L, Liang M, Hou X, Zhang Y, Zhang H, Guo Z, et al. The role of microRNA-16 in the pathogenesis of autoimmune diseases: a comprehensive review. Biomed Pharmacother. 2019;112:108583. [DOI] [PubMed] [Google Scholar]
- 51.Mohebi N, Damavandi E, Rostamian A-R, Javadi-Arjmand M, Movassaghi S, Choobineh H, et al. Comparison of plasma levels of MicroRNA-155-5p, MicroRNA-210-3p, and MicroRNA-16-5p in rheumatoid arthritis patients with healthy controls in a case-control study. Iran J Allergy Asthma Immunol. 2023;22:354–65. [DOI] [PubMed] [Google Scholar]
- 52.Wang M, Wang L, Zhang X, Yang X, Li X, Xia Q, et al. Overexpression of miR-31 in Peripheral Blood mononuclear cells (PBMC) from patients with Ankylosing Spondylitis. Med Sci Monit. 2017;23:5488–94. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 53.Schönauen K, Le N, von Arnim U, Schulz C, Malfertheiner P, Link A. Circulating and fecal microRNAs as biomarkers for inflammatory Bowel diseases. Inflamm Bowel Dis. 2018;24:1547–57. [DOI] [PubMed] [Google Scholar]
- 54.Nakano S, Oki M, Kusaka H. The role of p62/SQSTM1 in sporadic inclusion body myositis. Neuromuscul Disord. 2017;27:363–9. [DOI] [PubMed] [Google Scholar]
- 55.Milisenda JC, Pinal-Fernandez I, Lloyd TE, Grau JM, Miller FW, Selva-O’Callaghan A, et al. Accumulation of autophagosome cargo protein p62 is common in idiopathic inflammatory myopathies. Clin Exp Rheumatol. 2021;39:351–6. [PMC free article] [PubMed] [Google Scholar]
- 56.Selbach M, Schwanhäusser B, Thierfelder N, Fang Z, Khanin R, Rajewsky N. Widespread changes in protein synthesis induced by microRNAs. Nature. 2008;455:58–63. [DOI] [PubMed] [Google Scholar]
- 57.Amici DR, Pinal-Fernandez I, Mázala DAG, Lloyd TE, Corse AM, Christopher-Stine L, et al. Calcium dysregulation, functional calpainopathy, and endoplasmic reticulum stress in sporadic inclusion body myositis. Acta Neuropathol Commun. 2017;5:24. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 58.Ito S, Akamatsu Y, Noma A, Kimura S, Miyauchi K, Ikeuchi Y, et al. A single acetylation of 18 S rRNA is essential for Biogenesis of the small ribosomal subunit in Saccharomyces cerevisiae*. J Biol Chem. 2014;289:26201–12. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 59.Cortese A, Plagnol V, Brady S, Simone R, Lashley T, Acevedo-Arozena A, et al. Widespread RNA metabolism impairment in sporadic inclusion body myositis TDP43-proteinopathy. Neurobiol Aging. 2014;35:1491–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 60.Güttsches A-K, Brady S, Krause K, Maerkens A, Uszkoreit J, Eisenacher M, et al. Proteomics of rimmed vacuoles define new risk allele in inclusion body myositis. Ann Neurol. 2017;81:227–39. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 61.Sadowski CL, Henry RW, Lobo SM, Hernandez N. Targeting TBP to a non-TATA box cis-regulatory element: a TBP-containing complex activates transcription from snRNA promoters through the PSE. Genes Dev. 1993;7:1535–48. [DOI] [PubMed] [Google Scholar]
- 62.Kastner B, Will CL, Stark H, Lührmann R. Structural insights into Nuclear pre-mRNA splicing in higher eukaryotes. Cold Spring Harb Perspect Biol. 2019;11:a032417. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 63.Ma X, Liu Y, Zhang H, Qiu R, Zhao H, Xin Q, et al. Evidence for Genetic Association of CARD9 and SNAPC4 with Ankylosing spondylitis in a Chinese Han Population. J Rheumatol. 2014;41:318–24. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Additional file 1: Differential gene expression. Table with results of differential gene expression analysis.
Additional file 2: Differential miRNA expression. Table with results of miRNA expression analysis.
Additional file 3: Subnetwork nodes. Table of all nodes in the subnetworks, with their p-values.
Additional file 4: Subnetwork edges. Table of all edges in the subnetwork, with their score or correlation.
Additional file 5: Cell type correlations. Figure with correlations of subnetwork expression to estimated cell type proportions.
Data Availability Statement
The Workflow and code is available on GitHub (https://github.com/dwijnbergen/IBM_ASI_workflow) and WorkflowHub (10.48546/workflowhub.workflow.681.7). The Docker image is available on https://hub.docker.com/R/jdwijnbergen/multi-omics_asi, and it’s build requirements are available on Zenodo (10.5281/zenodo.10210364). The input data of our workflow is available on Zenodo at 10.5281/zenodo.10411125.