Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2024 Nov 1.
Published in final edited form as: J Autoimmun. 2023 Sep 27;140:103115. doi: 10.1016/j.jaut.2023.103115

Prioritization of Infectious Epitopes for Translational Investigation in Type 1 Diabetes Etiology

Sejal Mistry 1,2, Ramkiran Gouripeddi 1,2,3, Julio C Facelli 1,2,3
PMCID: PMC10965504  NIHMSID: NIHMS1935718  PMID: 37774556

Abstract

Molecular mimicry is one mechanism by which infectious agents are thought to trigger islet autoimmunity in type 1 diabetes. With a growing number of reported infectious agents and islet antigens, strategies to prioritize the study of infectious agents are critically needed to expedite translational research into the etiology of type 1 diabetes. In this work, we developed an in-silico pipeline for assessing molecular mimicry in type 1 diabetes etiology based on sequence homology, empirical binding affinity to specific MHC molecules, and empirical potential for T-cell immunogenicity. We then assess whether potential molecular mimics were conserved across other pathogens known to infect humans. Overall, we identified 61 potentially high-impact molecular mimics showing sequence homology, strong empirical binding affinity, and empirical immunogenicity linked with specific MHC molecules. We further found that peptide sequences from 32 of these potential molecular mimics were conserved across several human pathogens. These findings facilitate translational evaluation of molecular mimicry in type 1 diabetes etiology by providing a curated and prioritized list of peptides from infectious agents for etiopathologic investigation. These results may also provide evidence for generation of infectious and HLA-specific preclinical models and inform future screening and preventative efforts in genetically susceptible populations.

Keywords: Type 1 diabetes mellitus, islet autoimmunity, molecular mimicry, bioinformatics, translational science, sequence homology

Graphical Abstract

graphic file with name nihms-1935718-f0001.jpg

1. INTRODUCTION

A growing body of evidence supports an important role for infectious agents in the onset of type 1 diabetes etiology (1, 2). One proposed mechanism by which infectious agents trigger autoimmunity is molecular mimicry (3, 4). Molecular mimicry in type 1 diabetes may occur when infectious agents containing epitopes that resemble portions of certain proteins in pancreatic β-cells illicit cross-reactive immune responses against both the infectious agent and self-antigens in the β-cells (5, 6). While molecular structure and electrostatic interactions will ultimately define molecular mimicry (7), sequence homology is a powerful indicator of the likelihood of a given infectious agent to induce molecular mimicry. Previous studies have identified sequence homology between glutamic acid decarboxylase 65 (GAD65), a neuroendocrine enzyme detected in β-cells, and cytomegalovirus (8), rotavirus (9), coxsackie virus (10). Similarly, sequence homology between insulinoma-associated protein 2 (IA-2), a tyrosine phosphatase involved in neuroendocrine secretory granules, and rotavirus (9), enterovirus (11) have also been identified.

Despite these findings, infectious agents causative of type 1 diabetes have not been widely identified. Previous studies focus on identifying sequence homology between single or few infectious agents with a small subset of islet antigens implicated in type 1 diabetes (8-12). Translational evaluation of sequence homology is difficult, time-consuming, and costly because rigorous investigation requires assessment of cross-reactivity, determination of functional consequence, reproducibility in preclinical models, and collection of epidemiologic evidence (13). With over 1,400 infectious agents implicated in human disease (14, 15) and 854 epitopes implicated in type 1 diabetes (16), there are more than one million combinations of possible mimicking peptides. Traditional approaches to evaluate molecular mimicry at scale are not feasible. Therefore, strategies to prioritize infectious agents for translational evaluation are critically needed. Such an approach will expedite translational research into the etiology of type 1 diabetes by identifying infectious targets for mechanistic evaluation, provide evidence for generating infectious and HLA-specific preclinical models, and inform future studies for screening and preventative efforts in genetically susceptible populations.

In-silico approaches for molecular mimicry may provide scalable methods for prioritizing peptides from infectious agents for translational evaluation. Previous computational approaches have examined sequence homology between proteins from helminths and human proteins (17), specific pathogenic and non-pathogenic bacterial proteins with reference human proteins (18), and specific strains of clostridium botulinum and human proteins (19). These approaches were largely limited to specific pathogens and broadly considered many human proteins not explicitly implicated in autoimmune diseases. Few studies computationally evaluated molecular mimicry beyond sequence homology, resulting in thousands of implicated peptides without prioritization of pathogenicity. Biologically informed strategies to prioritize peptides that consider multiple classes of infectious agents may provide novel insights into the etiopathology of type 1 diabetes and other autoimmune diseases.

This work aimed to identify potentially novel infectious agents in type 1 diabetes etiology by developing an in-silico pipeline for assessing molecular mimicry. We generate a prioritized list of established infectious epitopes demonstrating 1.) sequence homology, 2.) similar strong empirical binding affinity to specific MHC molecules, and 3.) similar strong empirical potential for T-cell immunogenicity as known type 1 diabetes epitopes. We then characterize important biological and structural features of epitopes in this prioritized list. Finally, we determine if infectious epitopes in this prioritized list are conserved among other species of human pathogens. Together, this work provides novel insights into potential targets for molecular mimicry in type 1 diabetes etiology.

2. MATERIALS AND METHODS

2.1. Data Sources:

2.1.1. Type 1 diabetes epitopes:

Type 1 diabetes epitopes (ET1D) were defined as peptides isolated from specific pancreatic β-cell proteins recognized by T-cells that were identified in previously published experimental literature and cataloged in the Immune Epitope Database (IEDB). ET1D were previously systematically reviewed, cross-indexed with IEDB identifiers, and graded according to the degree of evidence available (16). Conventional CD4+ and CD8+ T-cell ET1D identified from human samples with evidence of natural processing or presentation were included, and duplicated epitopes were excluded from analysis. The IEDB identifier, sequence, antigen name and UniProt identifier, type of T-cell generated (CD4+ or CD8+), type of MHC receptor the epitope was presented by, and evidence grade were extracted for each ET1D (Supplementary Table 1).

2.1.2. Infectious epitopes:

Infectious epitopes (EINF) were defined as peptides isolated from specific infectious agent proteins recognized by T-cells identified in previously published experimental literature and cataloged in IEDB. IEDB was queried on December 14th, 2022 for epitopes from infectious diseases with a human host, T-cell or MHC ligand assays, and cited in peer-reviewed literature (20). The IEDB identifier, sequence, antigen name and UniProt identifier, and the organism’s name and NCBI taxonomy were extracted for each EINF. EINF with missing data, gaps in the epitope sequence, or epitopes originating from organisms that do not infect humans were excluded from analysis (Supplementary Table 2).

2.2. In-Silico Pipeline for Assessing Molecular Mimicry Between EINF and ET1D:

2.2.1. Sequence Homology Search:

The potential for molecular mimicry was first assessed by identifying ET1D that shared sequence homology with EINF. For each ET1D, sequence homology to every EINF was assessed using BlastP with the short task algorithm and default parameters (21).

Unique pairs of the 1,064,799 possible combinations of ET1D and EINF were considered to demonstrate sequence homology (SeqHT1D-INF) if the length of the identified homologous peptides in the BlastP hit was ≥ 5 amino acids and contained no gaps, The 6,247 SeqHT1D-INF meeting these criteria were eligible for binding affinity assessment.

2.2.2. Empirical Binding Affinity Calculation:

Molecular mimicry was further assessed by determining if the EINF that shared sequence homology with a ET1D could bind with strong empirical affinity to the same MHC molecule. Affinity binding predictions require a target MHC molecule, epitope, T-cell type, and prediction method. First, we identified the target MHC molecule for empirical binding affinity prediction. For each SeqHT1D-INF, the target MHC molecule was selected based on the experimentally observed haplotype associated with the ET1D reported in previous literature (16). MHC molecules could be reported in complete or incomplete HLA notation. Complete MHC molecules included the HLA gene, allele group, and specific HLA protein, noted as “HLA-(gene)*(allele group):(specific HLA protein)”. Incomplete MHC molecules lacked the allele group and/or specific HLA protein. If the complete MHC molecules definition was available, the specified molecule was used for affinity calculation. If an incomplete MHC molecule was provided, empirical binding affinity was calculated for each of the frequent MHC molecules from the general population (22, 23) (Supplementary Table 3) that contained the same allele group and/or specific HLA protein as the incomplete MHC molecule.

Next, the length of the ET1D and EINF within each SeqHT1D-INF was determined. The accepted peptide length is 8 – 14 amino acids for MHC-I molecules and 13 – 25 for MHC-II molecules (24, 25). SeqHT1D-INF with epitopes shorter than the lower limit of the accepted peptide length were excluded from the predicted binding affinity analysis. For SeqHT1D-INF with epitopes longer than the upper limit of the accepted peptide length, the upper limit was used as the epitope length. For SeqHT1D-INF with epitopes within the accepted peptide length, the direct epitope length was used. Then, we selected the appropriate MHC tool for analysis for each SeqHT1D-INF. If the MHC molecule identified in the above steps was associated with CD4+ T-cells, the MHC-I binding prediction tool (http://tools.iedb.org/mhci/) was used. If the MHC molecule identified was associated with CD8+ T-cells, the MHC-II binding prediction tool (http://tools.iedb.org/mhcii/) was used. Finally, the binding affinity algorithm was selected. The IEDB recommended 2020.09 (NetMHCpan EL 4.1) (26-30) and IEDB recommended 2.22 (27, 31, 32) prediction methods were used for MHC-I and MHC-II binding prediction tools, respectively.

For each SeqHT1D-INF, empirical binding affinity was first calculated between ET1D and its associated MHC molecule using the procedure described above. If multiple calculations were performed due to incomplete MHC molecules, the MHC molecule with the highest predicted binding affinity determined by the percentile rank was selected as the associated MHC molecule. Empirical binding affinity was then calculated between EINF and the MHC molecule associated with ET1D. The IC50 score, percentile rank, and core peptide recognized by the MHC molecule were reported for each calculation. ET1D and EINF for each SeqHT1D-INF were considered strong binders to specific MHC molecules (StrongBT1D-INF) if 1.) the percentile rank was ≥1% for MHC-I molecules and ≥10% for MHC-II molecules (24), respectively, and 2.) the predicted core binding peptides recognized by the MHC molecule for the ET1D and EINF were contained within the homologous peptides of SeqHT1D-INF or vice versa. The 216 pairs of StrongHT1D-INF meeting these criteria were included for immunogenicity analysis.

2.2.3. Empirical Immunogenicity Assessment:

Potential molecular mimics with sequence homology and strong empirical binding affinity were prioritized by empirically assessing immunogenicity for CD4+ or CD8+ T-cells. Immunogenicity for MHC-I molecules was assessed using a Class I pMHC Immunogenicity Predictor (https://nextgen-tools.iedb.org/tc1) (33). For MHC-II molecules, immunogenicity was assessed using Class II Immunogenicity Predictor (http://tools.iedb.org/CD4episcore/) (34). This tool is only validated for epitopes ≥ 15 peptides in length and for 7 alleles (35). Therefore, epitopes < 15 peptides in length and epitopes associated with ineligible MHC-II molecules (HLA-DRB1*04:01 and HLA-DQA1*03:01/ DQB1*03:02) were excluded. For each StrongBT1D-INF, an immunogenicity score was first calculated between the ET1D and its associated MHC molecule. Next, an immunogenicity score was calculated between EINF and the same MHC molecule associated with ET1D. For MHC-I molecules, the immunogenicity score represents the likelihood of the peptide/allele combination resulting in an immune response, where a positive number indicates an increased likelihood for immunogenicity. For MHC-II molecules, the immunogenicity score ranges from 0 – 100, with lower values indicating higher immunogenic responses. ET1D and EINF that were StrongBT1D-INF to a specific MHC molecule were considered to illicit empirical immunogenic responses (ImmunoRT1D-INF) if the propensity score was > 0 for MHC-I molecules (33) and < 50 for MHC-II molecules (34, 35). Ultimately, this resulted in 61 ImmunoRT1D-INF candidates for descriptive analysis.

2.3. Characterization of Prioritized Potential Molecular Mimics:

ET1D and EINF meeting the criteria of SeqHT1D-INF, StrongBT1D-INF, and ImmunoRT1D-INF were considered likely to be potential molecular mimics and were prioritized by the immunogenicity score of EINF. For each potential molecular mimic, the relationship between the antigen from ET1D, organism from EINF, and affiliated MHC molecule was visualized using a chord diagram (MNE-Connectivity, v0.06).

Next, the conformation of peptides implicated in potential molecular mimics of ET1D were visualized. AlphaFold structure predictions for each islet antigen were downloaded (36, 37) and visualizations were rendered with ChimeraX (38). Homologous peptides of ET1D identified through BlastP search that were completely contained in other homologous peptides within 3 amino acids of one another were merged. The structural conformations of these merged homologous peptides were highlighted on the AlphaFold structure predictions for each islet antigen. To gain further insight into the conformation of merged homologous peptides, the hydrophobicity/hydrophilicity of merged peptide domains was evaluated using the Hydrophobicity/ Hydrophilicity Analysis Tool (https://www.peptide2.com/N_peptide_hydrophobicity_hydrophilicity.php).

Finally, the biological, cellular, and molecular functions of proteins from EINF in potential molecular mimics were characterized. Gene Ontology (GO) terms (39) were extracted using the UniProt ID Mapping tool (40). GO terms were summarized into clusters using multidimensional scaling with 0.5 filtering and default similarity measures in REVIGO (41). Clusters were visualized using tree maps (plotly, v5.14.0).

2.4. Identification of Conserved Domains:

To determine if homologous domains from potential molecular mimics were conserved across other species for which epitope information was not available, we performed a peptide search for each homologous domain from EINF and pathogens known to infect humans. Known human pathogens were extracted from previous literature surveys conducted in 2001 and 2005 (14, 15). To identify additional novel pathogens, we queried the Emerging Infectious Disease Journal (https://wwwnc.cdc.gov/eid/AdvancedSearch) for articles published after 2005 with the search term “novel human.” The title and abstract of all entries were manually reviewed and novel human pathogens not identified in the previous literature were included for analysis. Human pathogens with NCBI taxonomy codes were considered for analysis.

For each human pathogen, all non-redundant protein sequences using UniProt reference clusters (UniRef) of protein sequences at 90% sequence identity were downloaded on December 20th, 2022 (42). Any fragment, hypothetical, or uncharacterized proteins were excluded from analysis. Peptide searches were performed between every included protein for each human pathogen and each homologous domain from EINF of potential molecular mimics. Comparisons between human pathogens and potential molecular mimics from the same organism were not considered. Homologous peptides between EINF of potential molecular mimics and human pathogen proteins were considered conserved domains (ConservedDINF-HPP) if both peptides shared 100% homology and the full peptide from EINF of potential molecular mimics was contained within the human pathogen protein. ConservedDINF-HPP were summarized by the number of organisms and proteins from human pathogens sharing the peptide sequence. The percentages of unique organisms from human pathogens sharing taxonomy with the infectious organism from which EINF of ConservedDINF-HPP originated were also summarized.

2.5. Data and Resource Availability:

Input data for this study are available through IEDB (www.iedb.org) (20) and through previous literature (16) and are reported in Supplementary Table 1 - 2. All analyses were performed in Python (v3.9.12). The BLAST algorithm can be accessed through The National Center for Biotechnology Information (https://blast.ncbi.nlm.nih.gov/Blast.cgi). MHC binding and immunogenicity tools can be accepted through IEDB Analysis Resources (http://tools.iedb.org/main/) (24).

3. RESULTS

3.1. Identification of Potential Molecular Mimicry Between ET1D and EINF:

3.1.1. Characterization of Epitope Data Sources:

Of the 854 systematically reviewed ET1D extracted from previous literature (16), 741 were excluded because they were not established in human samples, lacked antigen presentation, or were duplicated. Ultimately, 113 unique ET1D were included across islet antigens of insulin (INS), receptor-type tyrosine-protein phosphatase-like N (IA-2), glutamate decarboxylase 2 (GAD65), Zinc transporter 8 (ZnT8), protein S100-B (S100B), kinesin-like protein KIF1A (KIF1A), potassium channel subfamily k member 16 (KCNK16), neuroendocrine convertase 2 (PC2), insulin gene enhancer protein ISL-1 (ISL-1), chromogranin-A (CHGA), urocortin-3 (UCN3), and neuroendocrine protein 7B2 (SCG5) (Fig. 1A). Included ET1D were primarily associated with CD8+ T-cells (n = 43, 61.08%, Fig. 2A) and the most represented islet antigens were GAD65, INS, ZnT8, and IA-2 (Fig. 2B). Of the 12,740 EINF extracted from IEDB, 2,810 were excluded due to missing epitope data, sequence gaps, or because they were isolated in non-human pathogens. Ultimately, 9,330 unique EINF were included (Fig. 1A). These EINF were attributed to 114 different organisms, with epitopes from SARS-CoV-2, Hepacivirus C, and Mycobacterium tuberculosis being highly represented (Supplementary Fig. 1).

Fig. 1. Summary of potential molecular mimics identified through in-silico pipeline:

Fig. 1.

Open in a new tab

Overview of pipeline steps, inclusion/exclusion criteria, and results. (A) Data sources and inclusion/exclusion criteria for T1DM epitopes (ET1D) and infectious epitopes (EINF). (B) Sequence homology results and inclusion/exclusion criteria to identify unique pairs of ET1D and EINF that demonstrate sequence homology (SeqHT1D-INF). (C) Empirical binding affinity calculations and inclusion/exclusion criteria to identify ET1D and EINF for each SeqHT1D-INF that strongly bind to specific MHC molecules (StrongBT1D-INF). (D) Empirical immunogenicity assessment and inclusion/exclusion criteria to identify ET1D and EINF that were StrongBT1D-INF to a specific MHC molecule that illicit immunogenic responses (ImmunoRT1D-INF). (E) Identification of conserved domains and inclusion/exclusion criteria of human pathogens using a peptide search.

Fig. 2. Characteristics of potential molecular mimics by pipeline steps:

Fig. 2.

Open in a new tab

For pipeline steps of data source (pink), sequence homology search (orange), empirical binding affinity calculations (green), and empirical immunogenicity assessments (blue), bar graphs of the percentage of epitopes associated with (A) a specific T-cell type (CD4+, CD8+, or both), and (B) antigen from the type 1 diabetes epitope.

3.1.2. Summary of Sequence Homology Searches:

To identify potential targets for molecular mimicry, sequence homology searches were performed between each ET1D and EINF using BlastP. Among the 1,053,290 BlastP searches performed, sequence homology was identified between 7,787 unique ET1D and EINF (Fig. 1B). Among these BlastP hits, 1,540 were excluded due to small peptide size or presence of gaps, resulting in 6,247 SeqHT1D-INF for further analysis (0.59% hit rate) (Fig. 1B). SeqHT1D-INF contained more ET1D against CD4+ T-cells than CD8+ T cells (n = 3,798, 60.8% and n = 2,435, 38.82%, respectively, Fig. 2A) and from GAD65, INS, and IA-2 antigens (Fig. 2B). EINF for SeqHT1D-INF were distributed across similarly across infectious agents as the data sources (Supplementary Fig. 1).

3.1.3. Summary of Empirical Binding Affinity Calculations:

To assess biological plausibility of potential targets for molecular mimicry, empirical binding affinity was calculated between a specific MHC molecule and the ET1D and EINF for each SeqHT1D-INF. Among the 6,247 SeqHT1D-INF, empirical binding affinity could not be assessed for 767 SeqHT1D-INF due to ineligible epitope lengths (Fig. 1C). Of the remaining 5,480 SeqHT1D-INF, 5,034 demonstrated low predicted binding affinity and 230 did not contain a core binding domain with the homologous peptides, resulting in 216 StrongBT1D-INF (3.94% hit rate) (Fig. 1C). More StrongBT1D-INF were specific to CD8+ T-cells than CD4+ T-cells (Fig. 2A) and INS was the most common antigen (Fig. 2B).

3.1.4. Summary of Empirical Immunogenicity Assessments:

To prioritize potential targets for molecular mimicry for translational evaluation, immunogenicity was evaluated between a specific MHC molecule and the ET1D and EINF for each StrongBT1D-INF. Among the 216 StrongHT1D-INF, immunogenicity could not be assessed for 33 StrongBT1D-INF due to ineligible MHC molecules or epitope length (Fig. 1D). Of the remaining 183 StrongBT1D-INF, 122 demonstrated low immunogenicity scores, resulting in 61 ImmunoRT1D-INF (33.33% hit rate) (Fig. 1D). CD8+ T-cells and INS continued to be the most common T-cell type and antigens for ImmunoRT1D-INF (Fig. 2A-B) with similar trends observed in infectious organisms as previous pipeline steps (Supplementary Fig. 1).

3.2. Characterization of Prioritized Potential Molecular Mimicry Epitopes:

Overall, 61 pairs of ET1D and EINF meeting the criteria of SeqHT1D-INF, StrongBT1D-INF, and ImmunoRT1D-INF were considered potential molecular mimics (Supplementary Table 4). This prioritized list encompassed ET1D against 9 unique islets antigens, EINF from 18 unique organisms, and were specific to 4 unique MHC molecules. Of the 61 potential molecular mimics, the majority were affiliated with only HLA-A*02:01 (n = 50 potential molecular mimics, 81.97%). These 50 potential molecular mimics were distributed across all 9 islet antigens and 14 infectious organisms (Fig. 3), suggesting that HLA-A*02:01 mediates empiric immunogenic responses broadly against several islet antigens. In contrast, the remaining 11 molecular mimics affiliated with other MHC molecules were specific to certain islet antigens and infectious organisms. Potential molecular mimics affiliated with HLA-B*08:01 were specific to INS (n = 2 potential molecular mimics, 3.28%) and demonstrated homology to proteins from human betaherpesvirus 6B and hepacivirus C (Fig. 3). Potential molecular mimics affiliated with HLA-DRB3*02:02 were specific to IA-2 (n = 4 potential molecular mimics, 6.56%) and demonstrated homology to proteins from mycobacterium tuberculosis, leishmania brazilliensis, and coxiella burnetti (Fig. 3). Potential molecular mimics affiliated with HLA-DRB4*01:01 were specific to GAD65 (n = 2, 3.28%) and demonstrated homology to proteins from mycobacterium tuberculosis and human orthopneumovirus (Fig. 3). Potential molecular mimics affiliated with both HLA-A*02:01 and HLA-DRBB3*02:02 demonstrated sequence homology between IA-2 and hepacivirus C (n = 3 potential molecular mimics, 4.92%) (Fig. 3). Overall, these findings demonstrate MHC molecule, antigen, and organism-specific patterns of potential molecular mimicry.

Fig. 3. Distribution of potential molecular mimics by islet antigen, infectious organism, and MHC molecules:

Fig. 3.

Open in a new tab

Chord diagram demonstrating the relationship among potential molecular mimics. The top nodes represent antigens from the type 1 diabetes epitopes (red – yellow color scale) and the bottom nodes represent organisms from the infectious epitopes (green – blue color scale). The color of the chords (gray color scale) corresponds to the number of potential molecular mimics associated with the type 1 diabetes and infectious epitopes. The MHC molecule with which each potential molecular mimic is specified (purple color scale).

To evaluate the peptides implicated in potential molecular mimics on islet antigens, merged peptide domains on islet antigens from ET1D from prioritized molecular mimics were summarized (Supplementary Table 5) and the conformation of these peptides on their respective islet antigens were visualized (Fig. 4). Insulin contained 5 merged peptide domains, 4 of which were affiliated with HLA-A*02:01 and 1 was affiliated with HLA-B*08:01. GAD65 and IA-2 both contained 3 merged peptide domains affiliated with HLA-DRB4*01:01 or HLA-A*02:01. The remaining merged peptide domains were affiliated with HLA-A*02:01, with ZnT8 containing 3 merged peptide domains and S100B, KCNK16, PC2, ISL-1, and UCN3 all containing 1 merged peptide domain. Hydrophobicity/hydrophilicity analysis revealed that most of the ET1D homology sequences depicted in Fig. 4 were predominantly hydrophobic (Supplementary Table 5).

Fig. 4. Structural rendering of merged peptide domains from molecular mimics on islet antigens:

Fig. 4.

Open in a new tab

Merged peptide domains were highlighted on conformational renderings of islet antigen proteins. Colors of merged peptide domains correspond to the MHC molecule with which the molecular mimic is associated (red: HLA-A*02:01; green: HLA-B*08:01; blue: HLA-DRB3*02:02; pink: HLA-DRB4*01:01).

To characterize the biological, cellular, and molecular functions of proteins from which the EINF in potential molecular mimics were extracted, their GO terms were clustered and visualized using tree maps (Supplementary Table 6). We identified 7 clusters of biological function across proteins from infectious epitopes of molecular mimics, with biological GO terms of viral RNA genome replication and positive regulation of RNA biosynthetic processes being highly represented (Supplementary Fig. 2A). Clusters of cellular component GO terms resulted in 8 clusters, with endopeptidase complex, membrane, and host cell nucleus being highly represented (Supplementary Fig. 2B). We also identified 17 clusters of molecular function GO terms, with identical protein binding, RNA-dependent RNA polymerase activity, and ATP hydrolysis activity being highly represented (Supplementary Fig. 2C).

3.3. Conserved Domains Across Other Human Pathogens:

To determine if homologous domains from potential molecular mimics were conserved across other species for which epitope information was not available, we performed a peptide search for ConservedDINF-HPP.

Overall, 1,419 sequences of human pathogen proteins were extracted from a previous literature survey conducted before 2005 and 11 additional novel human pathogens were identified through an updated literature search (Fig. 1E). Eligible UniRef proteins and NCBI taxonomy codes were available for the proteins of 1,022 human pathogens (Supplementary Table 7). Of the 61 potential molecular mimics, 32 (52.46%) were considered ConservedDINF-HPP after peptide search between each protein from included human pathogens and homologous domains of EINF of potential molecular mimics. These ConservedDINF-HPP represented 30,816 UniRef proteins from 687 distinct human pathogens.

ConservedDINF-HPP containing the highest number of organisms and proteins included the pattern LLGLL originating from EINF of mycobacterium tuberculosis (635 unique organisms, 13,327 unique UniRef proteins), RRLAE from human betaherpesvirus 6B (516 unique organisms, 7,277 unique UniRef proteins), and ALLAF from leishmania major (545 unique organisms, 5,149 unique UniRef proteins) (Fig. 5, left), suggesting that these peptides are highly conserved across several other human pathogens. Majority of organisms from which ConservedDINF-HPP originated did not share any taxonomy lineages with the human pathogens that shared the homologous domains (Fig. 5, right), suggesting that these conserved domains are not specific to a certain taxonomy.

Fig 5. Distribution of conserved domains by human pathogens, human pathogen proteins, and taxonomy:

Fig 5.

Open in a new tab

Left: Conserved domains were first summarized by the number of unique human pathogens and the number of unique human pathogen proteins (size of marker). Conserved domains are grouped and colored by the infectious organism from which the epitope originated. Right: For each infectious organism from which the conserved domain originated, the percentage of human pathogens sharing taxonomy lineages were calculated by genus, family, order, class, phylum, and kingdom. Calculations are additive, such that the percentage of a higher taxonomic lineage includes all percentages of a lower taxonomic lineage.

4. DISCUSSION

In this study, we leverage state-of-the-art bioinformatics techniques with validated immunologic repository data to develop an in-silico pipeline for assessing molecular mimicry in type 1 diabetes etiology. We identified 61 potential molecular mimics showing sequence homology, strong empirical binding affinity, and empirical immunogenicity linked with specific MHC molecules (Supplementary Table 4). We further found that 32 peptide sequences were conserved across multiple human pathogens. Together, these findings provide a prioritized list of potential molecular mimics for further translational evaluation, including molecular structure and dynamics studies.

Peptides from potential molecular mimics spanned several infectious agents previously implicated in type 1 diabetes etiology. Mumps virus, human alphaherpesvirus 3, and SARS-CoV-2 have been previously evaluated as viral triggers of type 1 diabetes (43-45) and additional epitopes from these agents were identified in our analysis. Notably, our work supports previous findings that epitopes from SARS-CoV-2 may produce cross-reactive responses with pancreatic β-cells proteins (46). Given reports of increased incidence in type 1 diabetes in children following COVID-19 infection (47), our findings suggest that type 1 diabetes may be an important post-COVID autoimmune sequalae and reinforce the importance of vaccination in youth. Future studies should seek to experimentally investigate epitopes from mumps, human alphaherpesvirus 3, and SARS-CoV-2 identified in this work, as infection with these diseases may be implicated in type 1 diabetes etiology.

This work identified one epitope from human betaherpesvirus 6B as a potential molecular mimic. Notably, we previously found that sixth disease, caused by human betaherpesvirus 6B, at 12 months of age significantly increased risk for multiple islet autoantibodies and onset of type 1 diabetes among children with increased genetic susceptibility for type 1 diabetes (48). Taken together, these findings provide clinical and computational evidence for human betaherpesvirus 6B in type 1 diabetes etiology that has not been reported in previous literature. Experimental studies are needed to validate these findings and investigate epitopes from human betaherpesvirus 6B in HLA-specific preclinical models for type 1 diabetes etiology.

We also identified infectious agents known to cause respiratory illnesses, including human orthopneumovirus and human mastadenovirus C, as potential molecular mimics. Respiratory infections have been previously implicated in islet autoimmunity and type 1 diabetes etiology (48, 49), but pathogen-specific causes have been difficult to ascertain. The identification of prominent causes of upper respiratory illnesses with molecular mimicry towards known islet antigens calls for further monitoring of these infections in individuals with genetic susceptibility to type 1 diabetes. This study also identified novel infectious agents with potential molecular mimicry to islet antigens that have not yet been investigated in type 1 diabetes etiology. Hepacivirus C, human alphaherpesvirus 1, hepatitis B virus, and human immunodeficiency virus 1, represent prevalent infectious agents that may be important to prioritize for further investigation.

The in-silico pipeline developed in this work can be adapted to identify other potential molecular mimics in autoimmune conditions with unclear etiologies. Molecular mimicry has been proposed to play a role in the development multiple sclerosis, rheumatoid arthritis, systemic lupus erythematosus, Sjogren’s disease, systemic sclerosis, autoimmune thyroiditis, and autoimmune hepatitis (50). While molecular mimicry is classically considered a mechanism by which infectious agents illicit cross-reactive immune responses, other environmental triggers of autoimmune diseases may leverage similar mechanisms. For example, beta-casein found in cow’s milk protein demonstrates similar structures as glucose transporter 2 that are expressed on pancreatic β-cells (51), and investigation of mimicry among dietary factors may provide additional insight into type 1 diabetes etiology. Taken together, the methodology presented in this work provide a new frontier for streamlined, cost-effective, and rigorous prioritization of environmental triggers for investigation in autoimmune disease etiology. There are several limitations to this work. First, this study relied on infectious epitope data available through IEDB. Though available epitopes were experimentally validated, the dearth of infectious agents was limited to well-studied organisms (20). To partially address this concern and provide broader insight into other pathogens that may not be highly represented in IEDB, we performed a peptide search among human pathogens identified in literature reviews. Our analytical pipeline addresses sequence homology and the empirical immunogenicity assessment performed in this work was only validated for a limited number of MHC II molecules (24). Future studies should seek to assess structural homology and evaluate a broader number of MHC molecules. Finally, the human pathogens selected for conservation analysis were provided by literature reviews conducted in 2001 and 2005 (14, 15) and may not represent novel pathogens discovered since then. To partially address this concern, we updated the pathogen list based on findings from Emerging Infectious Disease Journal, but some pathogens may still be missing.

5. CONCLUSIONS

This study provided the first comprehensive insights into potential molecular mimics of islet antigens implicated in type 1 diabetes using epitope data from experimentally validated immunologic repositories. Compared to previous approaches,, this study enhances the biological rigor of computational methods to assess molecular mimicry by combining sequence homology with empirical binding affinity and immunogenicity predictions. These findings facilitate translational evaluation of molecular mimicry in type 1 diabetes etiology by providing a curated and prioritized list of peptides from infectious. These results may also provide evidence for generation of infectious and HLA-specific preclinical models and inform future clinical studies for screening and vaccination efforts in genetically susceptible populations.

Supplementary Material

1
2

HIGHLIGHTS.

  • This work assesses molecular mimicry in type 1 diabetes using computational methods

  • We empirically evaluate sequence homology, binding affinity, and immunogenicity

  • This method identified 61 potential molecular mimics in type 1 diabetes etiology

  • 32 potential molecular mimics were conserved across other human pathogens

  • This pipeline can be adapted to investigate mimicry in other autoimmune diseases

Acknowledgements:

Computational resources for this study were provided by the Center for High Performance Computing at the University of Utah. Epitope information and analysis resources was downloaded through the IEDB (www.iedb.org) and IEDB Analysis Resource (http://tools.iedb.org/main/), respectively. Molecular graphics performed with UCSF ChimeraX, developed by the Resource for Biocomputing, Visualization, and Informatics at the University of California, San Francisco, with support from National Institutes of Health R01-GM129325 and the Office of Cyber Infrastructure and Computational Biology, National Institute of Allergy and Infectious Diseases.

Funding:

This work was supported by the National Institute of Diabetes and Digestive and Kidney Diseases (F30DK134113) and the National Center for Advancing Translational Sciences (UL1TR002538).

Footnotes

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Duality of Interest:

No potential conflicts of interest relevant to this article were reported.

RERFENCES

  • 1.Christen U, Bender C, von Herrath MG. Infection as a cause of type 1 diabetes? Curr Opin Rheumatol. 2012;24(4):417–23. doi: 10.1097/BOR.0b013e3283533719. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.Filippi CM, von Herrath MG. Viral trigger for type 1 diabetes: pros and cons. Diabetes. 2008;57(11):2863–71. doi: 10.2337/db07-1023. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Coppieters KT, Wiberg A, von Herrath MG. Viral infections and molecular mimicry in type 1 diabetes. Apmis. 2012;120(12):941–9. Epub 20121011. doi: 10.1111/apm.12011. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Cusick MF, Libbey JE, Fujinami RS. Molecular mimicry as a mechanism of autoimmune disease. Clin Rev Allergy Immunol. 2012;42(1):102–11. doi: 10.1007/s12016-011-8294-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Fujinami RS, Oldstone MB. Molecular mimicry as a mechanism for virus-induced autoimmunity. Immunol Res. 1989;8(1):3–15. doi: 10.1007/bf02918552. [DOI] [PubMed] [Google Scholar]
  • 6.Benoist C, Mathis D. Autoimmunity provoked by infection: how good is the case for T cell epitope mimicry? Nat Immunol. 2001;2(9):797–801. doi: 10.1038/ni0901-797. [DOI] [PubMed] [Google Scholar]
  • 7.Balbin CA, Nunez-Castilla J, Stebliankin V, Baral P, Sobhan M, Cickovski T, et al. Epitopedia: identifying molecular mimicry between pathogens and known immune epitopes. ImmunoInformatics. 2023;9:100023. doi: 10.1016/j.immuno.2023.100023. [DOI] [Google Scholar]
  • 8.Roep BO, Hiemstra HS, Schloot NC, De Vries RR, Chaudhuri A, Behan PO, et al. Molecular mimicry in type 1 diabetes: immune cross-reactivity between islet autoantigen and human cytomegalovirus but not Coxsackie virus. Ann N Y Acad Sci. 2002;958:163–5. [PubMed] [Google Scholar]
  • 9.Honeyman MC, Stone NL, Harrison LC. T-cell epitopes in type 1 diabetes autoantigen tyrosine phosphatase IA-2: potential for mimicry with rotavirus and other environmental agents. Mol Med. 1998;4(4):231–9. [PMC free article] [PubMed] [Google Scholar]
  • 10.Atkinson MA, Bowman MA, Campbell L, Darrow BL, Kaufman DL, Maclaren NK. Cellular immunity to a determinant common to glutamate decarboxylase and coxsackie virus in insulin-dependent diabetes. J Clin Invest. 1994;94(5):2125–9. doi: 10.1172/jci117567. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Härkönen T, Lankinen H, Davydova B, Hovi T, Roivainen M. Enterovirus infection can induce immune responses that cross-react with beta-cell autoantigen tyrosine phosphatase IA-2/IAR. J Med Virol. 2002;66(3):340–50. doi: 10.1002/jmv.2151. [DOI] [PubMed] [Google Scholar]
  • 12.Hiemstra HS, Schloot NC, van Veelen PA, Willemen SJ, Franken KL, van Rood JJ, et al. Cytomegalovirus in autoimmunity: T cell crossreactivity to viral antigen and autoantigen glutamic acid decarboxylase. Proc Natl Acad Sci U S A. 2001;98(7):3988–91. doi: 10.1073/pnas.071050898. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Peterson LK, Fujinami RS. 3 - MOLECULAR MIMICRY. In: Shoenfeld Y, Gershwin ME, Meroni PL, editors. Autoantibodies (Second Edition). Burlington: Elsevier; 2007. p. 13–9. [Google Scholar]
  • 14.Taylor LH, Latham SM, Woolhouse ME. Risk factors for human disease emergence. Philos Trans R Soc Lond B Biol Sci. 2001;356(1411):983–9. doi: 10.1098/rstb.2001.0888 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Woolhouse ME, Gowtage-Sequeria S. Host range and emerging and reemerging pathogens. Emerg Infect Dis. 2005;11(12):1842–7. doi: 10.3201/eid1112.050997. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.James EA, Mallone R, Kent SC, DiLorenzo TP. T-Cell Epitopes and Neo-epitopes in Type 1 Diabetes: A Comprehensive Update and Reappraisal. Diabetes. 2020;69(7):1311–35. doi: 10.2337/dbi19-0022. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Ludin P, Nilsson D, Mäser P. Genome-Wide Identification of Molecular Mimicry Candidates in Parasites. PLOS ONE. 2011;6(3):e17546. doi: 10.1371/journal.pone.0017546. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Doxey AC, McConkey BJ. Prediction of molecular mimicry candidates in human pathogenic bacteria. Virulence. 2013;4(6):453–66. doi: 10.4161/viru.25180. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Bhardwaj T, Haque S, Somvanshi P. In silico identification of molecular mimics involved in the pathogenesis of Clostridium botulinum ATCC 3502 strain. Microb Pathog. 2018;121:238–44. Epub 20180512. doi: 10.1016/j.micpath.2018.05.017. [DOI] [PubMed] [Google Scholar]
  • 20.Vita R, Mahajan S, Overton JA, Dhanda SK, Martini S, Cantrell JR, et al. The Immune Epitope Database (IEDB): 2018 update. Nucleic Acids Res. 2019;47(D1):D339–d43. doi: 10.1093/nar/gky1006. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Altschul SF, Gish W, Miller W, Myers EW, Lipman DJ. Basic local alignment search tool. J Mol Biol. 1990;215(3):403–10. doi: 10.1016/s0022-2836(05)80360-2. [DOI] [PubMed] [Google Scholar]
  • 22.Weiskopf D, Angelo MA, de Azeredo EL, Sidney J, Greenbaum JA, Fernando AN, et al. Comprehensive analysis of dengue virus-specific responses supports an HLA-linked protective role for CD8+ T cells. Proc Natl Acad Sci U S A. 2013;110(22):E2046–53. Epub 20130411. doi: 10.1073/pnas.1305227110. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Greenbaum J, Sidney J, Chung J, Brander C, Peters B, Sette A. Functional classification of class II human leukocyte antigen (HLA) molecules reveals seven different supertypes and a surprising degree of repertoire sharing across supertypes. Immunogenetics. 2011;63(6):325–35. Epub 20110209. doi: 10.1007/s00251-011-0513-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Fleri W, Paul S, Dhanda SK, Mahajan S, Xu X, Peters B, et al. The Immune Epitope Database and Analysis Resource in Epitope Discovery and Synthetic Vaccine Design. Front Immunol. 2017;8:278. Epub 20170314. doi: 10.3389/fimmu.2017.00278. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Paul S, Kolla RV, Sidney J, Weiskopf D, Fleri W, Kim Y, et al. Evaluating the Immunogenicity of Protein Drugs by Applying In Vitro MHC Binding Data and the Immune Epitope Database and Analysis Resource. Clinical and Developmental Immunology. 2013;2013:467852. doi: 10.1155/2013/467852. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Nielsen M, Lundegaard C, Worning P, Lauemøller SL, Lamberth K, Buus S, et al. Reliable prediction of T-cell epitopes using neural networks with novel sequence representations. Protein Sci. 2003;12(5):1007–17. doi: 10.1110/ps.0239403. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Bui HH, Sidney J, Peters B, Sathiamurthy M, Sinichi A, Purton KA, et al. Automated generation and evaluation of specific MHC binding predictive tools: ARB matrix applications. Immunogenetics. 2005;57(5):304–14. Epub 20050503. doi: 10.1007/s00251-005-0798-y. [DOI] [PubMed] [Google Scholar]
  • 28.Peters B, Sette A. Generating quantitative models describing the sequence specificity of biological processes with the stabilized matrix method. BMC Bioinformatics. 2005;6(1):132. doi: 10.1186/1471-2105-6-132. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Karosiene E, Lundegaard C, Lund O, Nielsen M. NetMHCcons: a consensus method for the major histocompatibility complex class I predictions. Immunogenetics. 2012;64(3):177–86. Epub 20111020. doi: 10.1007/s00251-011-0579-8 [DOI] [PubMed] [Google Scholar]
  • 30.Zhang H, Lund O, Nielsen M. The PickPocket method for predicting binding specificities for receptors based on receptor pocket similarities: application to MHC-peptide binding. Bioinformatics. 2009;25(10):1293–9. Epub 20090317. doi: 10.1093/bioinformatics/btp137. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Nielsen M, Lundegaard C, Lund O. Prediction of MHC class II binding affinity using SMM-align, a novel stabilization matrix alignment method. BMC Bioinformatics. 2007;8(1):238. doi: 10.1186/1471-2105-8-238. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Sturniolo T, Bono E, Ding J, Raddrizzani L, Tuereci O, Sahin U, et al. Generation of tissue-specific and promiscuous HLA ligand databases using DNA microarrays and virtual HLA class II matrices. Nat Biotechnol. 1999;17(6):555–61. doi: 10.1038/9858. [DOI] [PubMed] [Google Scholar]
  • 33.Calis JJ, Maybeno M, Greenbaum JA, Weiskopf D, De Silva AD, Sette A, et al. Properties of MHC class I presented peptides that enhance immunogenicity. PLoS Comput Biol. 2013;9(10):e1003266. Epub 20131024. doi: 10.1371/journal.pcbi.1003266 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34.Dhanda SK, Karosiene E, Edwards L, Grifoni A, Paul S, Andreatta M, et al. Predicting HLA CD4 Immunogenicity in Human Populations. Front Immunol. 2018;9:1369. Epub 20180614. doi: 10.3389/fimmu.2018.01369 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.Paul S, Lindestam Arlehamn CS, Scriba TJ, Dillon MBC, Oseroff C, Hinz D, et al. Development and validation of a broad scheme for prediction of HLA class II restricted T cell epitopes. Journal of Immunological Methods. 2015;422:28–34. doi: 10.1016/j.jim.2015.03.022. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36.Varadi M, Anyango S, Deshpande M, Nair S, Natassia C, Yordanova G, et al. AlphaFold Protein Structure Database: massively expanding the structural coverage of protein-sequence space with high-accuracy models. Nucleic Acids Research. 2022;50(D1):D439–D44. doi: 10.1093/nar/gkab1061. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37.Jumper J, Evans R, Pritzel A, Green T, Figurnov M, Ronneberger O, et al. Highly accurate protein structure prediction with AlphaFold. Nature. 2021;596(7873):583–9. doi: 10.1038/s41586-021-03819-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.Pettersen EF, Goddard TD, Huang CC, Meng EC, Couch GS, Croll TI, et al. UCSF ChimeraX: Structure visualization for researchers, educators, and developers. Protein Sci. 2021;30(1):70–82. Epub 20201022. doi: 10.1002/pro.3943. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39.Ashburner M, Ball CA, Blake JA, Botstein D, Butler H, Cherry JM, et al. Gene ontology: tool for the unification of biology. The Gene Ontology Consortium. Nat Genet. 2000;25(1):25–9. doi: 10.1038/75556. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40.Pundir S, Martin MJ, O'Donovan C. UniProt Tools. Curr Protoc Bioinformatics. 2016;53:1.29.1–1..15. Epub 20160324. doi: 10.1002/0471250953.bi0129s53. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 41.Supek F, Bošnjak M, Škunca N, Šmuc T. REVIGO Summarizes and Visualizes Long Lists of Gene Ontology Terms. PLOS ONE. 2011;6(7):e21800. doi: 10.1371/journal.pone.0021800. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42.Suzek BE, Wang Y, Huang H, McGarvey PB, Wu CH, the UniProt C. UniRef clusters: a comprehensive and scalable alternative for improving sequence similarity searches. Bioinformatics. 2015;31(6):926–32. doi: 10.1093/bioinformatics/btu739. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 43.Hyöty H, Leinikki P, Reunanen A, Ilonen J, Surcel HM, Rilva A, et al. Mumps infections in the etiology of type 1 (insulin-dependent) diabetes. Diabetes Res. 1988;9(3):111–6. [PubMed] [Google Scholar]
  • 44.Kendall EK, Olaker VR, Kaelber DC, Xu R, Davis PB. Association of SARS-CoV-2 Infection With New-Onset Type 1 Diabetes Among Pediatric Patients From 2020 to 2021. JAMA Network Open. 2022;5(9):e2233014–e. doi: 10.1001/jamanetworkopen.2022.33014. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 45.Meziane FZ, Dali-Sahi M, Dennouni-Medjati N, Boulenouar H, Kachekouche Y, Benslama Y, et al. Molecular mimicry between varicella, measles virus and Hsp60 in type 1 diabetes associated HLA-DR3/DR4 molecules. Diabetes & Metabolic Syndrome: Clinical Research & Reviews. 2020;14(6):1783–9. doi: 10.1016/j.dsx.2020.08.009. [DOI] [PubMed] [Google Scholar]
  • 46.Churilov LP, Normatov MG, Utekhin VJ. Molecular Mimicry between SARS-CoV-2 and Human Endocrinocytes: A Prerequisite of Post-COVID-19 Endocrine Autoimmunity? Pathophysiology. 2022;29(3):486–94. Epub 20220825. doi: 10.3390/pathophysiology29030039. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 47.Barrett CE, Koyama AK, Alvarez P, Chow W, Lundeen EA, Perrine CG, et al. Risk for Newly Diagnosed Diabetes >30 Days After SARS-CoV-2 Infection Among Persons Aged <18 Years - United States, March 1, 2020-June 28, 2021. MMWR Morb Mortal Wkly Rep. 2022;71(2):59–65. Epub 20220114. doi: 10.15585/mmwr.mm7102e2. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 48.Mistry S, Gouripeddi R, Raman V, Facelli JC. Sequential data mining of infection patterns as predictors for onset of type 1 diabetes in genetically at-risk individuals. Journal of Biomedical Informatics. 2023:104385. doi: 10.1016/j.jbi.2023.104385. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 49.Lönnrot M, Lynch KF, Elding Larsson H, Lernmark Å, Rewers MJ, Törn C, et al. Respiratory infections are temporally associated with initiation of type 1 diabetes autoimmunity: the TEDDY study. Diabetologia. 2017;60(10):1931–40. Epub 20170802. doi: 10.1007/s00125-017-4365-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 50.Rojas M, Restrepo-Jiménez P, Monsalve DM, Pacheco Y, Acosta-Ampudia Y, Ramírez-Santana C, et al. Molecular mimicry and autoimmunity. Journal of Autoimmunity. 2018;95:100–23. doi: 10.1016/j.jaut.2018.10.012. [DOI] [PubMed] [Google Scholar]
  • 51.Pozzilli P. Beta-casein in cow's milk: a major antigenic determinant for type 1 diabetes? J Endocrinol Invest. 1999;22(7):562–7. doi: 10.1007/bf03343610. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

1
NIHMS1935718-supplement-1.xlsx (637KB, xlsx)
2
NIHMS1935718-supplement-2.docx (636.9KB, docx)

RESOURCES