Дивергентные аутоиммунные генетические ландшафты у диабетиков и недиабетиков.

Background

Autoimmune diseases (AD) are a group of disorders in which the immune system mistakenly attacks the body’s own cells, tissues, or organs, resulting in chronic inflammation, tissue damage, and systemic dysfunction [ 1 ]. ADs are multifactorial disorders in which genetic, environmental, hormonal, and immunological factors collectively contribute to disease development [ 2 ]. ADs encompass diverse organ-specific and systemic disorders, including gastrointestinal, neurological, endocrine, musculoskeletal, and systemic autoimmune conditions such as rheumatoid arthritis and systemic lupus erythematosus [ 3 ]. Epidemiological studies indicate that approximately 3–5% of the global population is at risk of autoimmune diseases [ 4 ]. The burden of ADs is rising globally, with yearly increases in the worldwide incidence and prevalence estimated at 19.1% and 12.5%, respectively [ 5 ]. In the present scenario, nearly 100 autoimmune diseases have been identified, with systemic lupus erythematosus (SLE), rheumatoid arthritis (RA), and multiple sclerosis (MS), being among the most prevalent [ 6 ]. In parallel, type 2 diabetes mellitus (T2D) is an ongoing condition resulting from insulin resistance, along with increased glucose levels, which has become a serious health problem across the globe and is increasingly seen as involving dysfunction of the immune and inflammatory system, which might be related to autoimmunity [ 7 – 9 ]. The existence of common inflammatory mechanisms is now understood to be an important component in the development of various diseases, such as T2D and autoimmune diseases.

Type 2 diabetes mellitus constitutes nearly 90% of all cases of diabetes [ 7 ]. The rapid increase in the prevalence of type 2 diabetes mellitus, along with its high morbidity and mortality, has become an important health concern [ 10 ]. The onset of T2D is multifactorial and consists of a combination of genetic predisposition and modern lifestyle influences, such as unhealthy dietary patterns, physical inactivity, and media that support an increasingly obesogenic environment [ 11 ]. Beyond classical metabolic dysfunction, T2D is increasingly recognised to involve chronic low-grade inflammation and immune dysregulation, suggesting potential genetic overlap with autoimmune disorders [ 12 , 13 ].

Advances in genome-wide association studies (GWAS) have substantially improved understanding of the genetic architecture of complex diseases. It has enabled researchers to identify disease-associated variants across the genome [ 14 ] including those associated with both autoimmune diseases and metabolic disorders, facilitating the development of polygenic risk score (PRS) approaches to estimate cumulative genetic susceptibility [ 15 ]. GWAS enable the systematic analysis of genetic variants across the genome to identify loci associated with complex traits and diseases. These studies provide insights into disease mechanisms, genetic architecture, heritability, and potential clinical applications, including risk prediction and drug discovery [ 16 ]. PRS aggregate the effects of multiple disease-associated single-nucleotide polymorphisms (SNPs) to estimate an individual’s genetic predisposition to complex diseases, including autoimmune disorders, using GWAS-derived summary statistics [ 17 ]. Comparative analyses of PRS distributions allow evaluation of genetic risk differences between population groups [ 18 ], while pathway enrichment analyses of genes mapped from disease-associated variants offer functional insights into the biological and immunological mechanisms underlying observed risk patterns [ 19 ]. However, most PRS models have been derived from European populations, limiting their transferability to genetically diverse South Asian populations.

Despite growing interest in the genetic links between metabolic and immune-mediated disorders, differences in autoimmune disease susceptibility between diabetic and non-diabetic individuals remain insufficiently explored. To date, few studies have systematically compared autoimmune disease polygenic susceptibility between diabetic and non-diabetic individuals in Indian populations using PRS-based approaches. This deficiency has been addressed by conducting a systematic analysis of polygenic risk scores for various autoimmune diseases in both diabetic and non-diabetic people in an Indian cohort. Pathway enrichment analysis was used in order to determine the functional pathways that are involved with genetic variants related to the development of disease. Through PRS-based profiling and pathway analysis, the current study hopes to shed light on the common genetic architecture underlying metabolic and autoimmune conditions, while also characterising the predilection and genetic burden of autoimmune diseases within the premeditated population.

Methods

Data collection and curation

Genomic data and associated demographic and clinical information were obtained from the Gujarat Biotechnology Research Centre (GBRC), Gujarat [ 20 ]. The dataset comprised genotyping data for 804,934 single-nucleotide variants (SNVs) from 474 anonymised individuals genotyped using the Axiom Affymetrix genotyping array. The cohort included 99 diabetic and 375 non-diabetic individuals, and this dataset served as the target file for subsequent polygenic risk score (PRS) analyses.

The age range of participants was 15–90 years and the cohort included 140 females and 334 males. Individuals were categorized based on diabetic status annotation available in the metadata. Among the diabetic individuals, 32 individuals exclusively had diabetes, whereas the remaining 67 individuals had diabetes along with one or more comorbid conditions, including obesity, hypertension, thyroid disorders, and lung disorders. The non-diabetic group included individuals without any reported medical condition ( N  = 233), individuals with unknown comorbidity status ( N  = 37), and individuals with non-diabetic comorbidities ( N  = 105) (Table  1 ). Detailed medication history and longitudinal clinical information were not available within the anonymized dataset metadata.

Table 1 Cohort characteristics Variable Diabetic Non-diabetic Sample size 99 375 Age range 15–90 15–90 Exclusively diabetes 32 NA Diabetes + comorbidities 67 NA No reported condition NA 233 Unknown status NA 37 Other comorbidities NA 105

Cohort characteristics

We note here that while diabetic status information was available for all included individuals, detailed clinical diagnoses for autoimmune diseases such as rheumatoid arthritis (RA), systemic lupus erythematosus (SLE), psoriasis, or other autoimmune conditions were not available in the anonymised dataset metadata. Therefore, the present study was designed to compare polygenic susceptibility profiles for autoimmune diseases between diabetic and non-diabetic individuals rather than evaluate clinically confirmed autoimmune disease comorbidities.

GWAS summary statistics (base files) for 11 autoimmune diseases were obtained in Variant Call Format (VCF) from the IEU OpenGWAS database following strict inclusion and exclusion criteria [ 21 , 22 ]. The selected autoimmune diseases included autoimmune hepatitis, ILD related to systemic autoimmune diseases, rheumatoid arthritis, psoriatic arthritis, systemic lupus erythematosus, and others, as detailed in Table 2 .

Table 2 Genome-wide association study (GWAS) datasets used for polygenic risk score analysis. This table summarises the 11 autoimmune disorders included in the study, along with their corresponding IEU OpenGWAS identifiers, ancestry of the discovery cohorts, number of cases and controls, total number of SNPs analysed, and year of data release. These GWAS datasets served as base files for the calculation of polygenic risk scores in diabetic and non-diabetic individuals SI. No Trait IEU OpenGwas ID Ancestry Type Number of cases Number of controls Number of SNP Year 1 Psoriatic arthritis ieu-b-5116 European 5065 21,286 NA 2022 2 Hypothyroidism, strict autoimmune finn-b-E4_HYTHY_AI_STRICT European 22,997 175,475 16,380,353 2021 3 Other autoimmune haemolytic anaemias finn-b-D3_AIHA_OTHER European 127 218,396 16,380,466 2021 4 Autoimmune hyperthyroidism finn-b-AUTOIMMUNE_HYPERTHYROIDISM European 962 172,976 16,380,189 2021 5 Severe autoimmune type 2 diabetes ebi-a-GCST90026412 European 452 2744 5,376,535 2021 6 Autoimmune hepatitis ebi-a-GCST90018785 European 821 484,413 24,198,482 2021 7 ILD related to systemic autoimmune disease finn-b-ILD_SYST_AUTO European 411 196,986 16,380,411 2021 8 Type 1 diabetes ebi-a-GCST90018925 European 6447 451,248 24,182,422 2021 9 Systemic lupus erythematosus ebi-a-GCST90011866 East Asian 4222 8431 5,691,661 2021 10 Myasthenia gravis ebi-a-GCST90093061 European 1873 36,370 23,679,120 2022 11 Rheumatoid arthritis ebi-a-GCST90018690 East Asian 5348 173,268 12,454,695 2021

Genome-wide association study (GWAS) datasets used for polygenic risk score analysis. This table summarises the 11 autoimmune disorders included in the study, along with their corresponding IEU OpenGWAS identifiers, ancestry of the discovery cohorts, number of cases and controls, total number of SNPs analysed, and year of data release. These GWAS datasets served as base files for the calculation of polygenic risk scores in diabetic and non-diabetic individuals

Population structure analysis

To investigate population structure and assess genetic homogeneity within the Gujarati cohort, principal component analysis (PCA) was performed using genome-wide SNP data from the study samples merged with reference populations from the Genome-Asia 100 K [ 23 ]. Prior to PCA, standard genotype quality control procedures were applied, including filtering variants with high missingness and performing linkage disequilibrium (LD) pruning to retain a set of approximately independent SNPs. PCA was conducted using the --pca function implemented in PLINK v1.9 [ 24 ].

Polygenic Risk Score (PRS) estimation

Polygenic risk score (PRS) comparisons were performed for 11 biologically relevant autoimmune diseases selected a priori based on previously reported immune-metabolic associations in the literature. As the present study represents a targeted hypothesis-driven exploratory analysis rather than a genome-wide discovery framework, nominal p-values generated from PRS association analyses were interpreted cautiously. The analyses were intended to evaluate relative differences in autoimmune genetic susceptibility between diabetic and non-diabetic individuals rather than identify novel disease-associated loci.

GWAS summary statistics for all 11 autoimmune disorders were processed by extracting key parameters, including chromosome, SNP identifier, base position, effect allele (A1), non-effect allele (A2), beta value, standard error, and log P-value. Log-transformed P-values were converted to standard P-values prior to analysis. Variant extraction and preprocessing were performed using Linux-based computational pipelines (Ubuntu v24).

PRS calculations were performed using PRSice v2 [ 25 , 26 ] within a standard clumping-and-thresholding framework. Separate PRS models were generated for each of the 11 autoimmune diseases, and the best-performing model for each disease was selected based on model fit. We note here that the “best-performing model” refers to the PRS model showing the strongest statistical association and overall model fit within the examined threshold range.

Linkage disequilibrium (LD)-based clumping was performed using a 250 kb genomic window (--clump-kb 250) and an LD threshold of r² = 0.1 (--clump-r2 0.1) to retain relatively independent variants and reduce redundancy due to correlated SNPs. The clumping p-value threshold was set to 1.0 (--clump-p 1.0), so that all variants were initially considered during clumping before optimising the PRS thresholds. PRS models were then generated across a range of SNP inclusion p-value thresholds, from 5 × 10⁻⁸ (--lower 5e-08) to 0.5 (--upper 0.5), in increments of 5 × 10⁻⁵ (--interval 5e-05). These multiple-threshold models were evaluated by PRSice2 to identify the best-performing PRS model in the target dataset. For visualisation, bar-level thresholds included 0.001, 0.05, 0.1, 0.2, 0.3, 0.4, 0.5, and 1.0. Beta coefficients from the GWAS summary statistics (--stat BETA) were used as SNP effect sizes in PRS computation. The target genotype dataset comprised 22 autosomes (--num-auto 22) from the study cohort.

Statistical analysis

The present analysis primarily focused on comparative evaluation of autoimmune disease PRS distributions between diabetic and non-diabetic individuals. While cohort-level demographic and comorbidity information was available, comprehensive individual-level covariate data, including BMI measurements and uniformly structured clinical metadata, were not available for all individuals within the dataset. Therefore, fully adjusted regression analyses incorporating multiple clinical covariates could not be robustly implemented, and the findings should be interpreted within the context of an exploratory population-genetic analysis framework.

The best output files generated by PRSice2 for each autoimmune disease were stratified into diabetic and non-diabetic groups. Group-wise comparisons of PRS distributions were conducted using the non-parametric Wilcoxon rank sum test. A P-value threshold of < 0.05 was considered statistically significant.

Violin plots were generated to visualize PRS distributions, effect sizes, and group differences for selected autoimmune diseases. All statistical analyses and visualizations were performed using standard R v4.5.2 [ 27 ].

Pathway enrichment analysis

Pathway enrichment analysis was conducted for rheumatoid arthritis (RA) and systemic lupus erythematosus (SLE) to investigate the biological relevance of disease-associated candidate genetic variants. Given the exploratory nature of pathway enrichment analyses and the potential for inflated pathway signals due to inclusion of weakly associated loci, a conservative threshold was considered appropriate for identifying biologically relevant pathways. Therefore, for downstream pathway enrichment analysis, SNPs surpassing a significance threshold of P  < 1 × 10⁻⁸ were selected from the GWAS summary statistics. This more stringent threshold was intentionally applied to prioritise high-confidence disease-associated loci and reduce the inclusion of weak or potentially spurious associations in downstream functional interpretation. The stricter cutoff was chosen to improve the biological specificity and robustness of the enriched pathways identified during functional annotation analyses.

GWAS summary statistics were filtered based on stringent significance thresholds ( P < 1 × 10⁻⁸), and overlapping SNPs between the target genotype dataset and disease-specific GWAS datasets were identified. This resulted in 1,881 overlapping SNPs for RA and 625 for SLE. These SNPs were functionally annotated and mapped to genes using the SNPnexus platform. Annotated gene lists were subjected to Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis using ShinyGO (version 0.85.1), with a false discovery rate (FDR) threshold of 0.01 [ 28 , 29 ]. Enriched pathways were visualised using lollipop plots and KEGG pathway maps.

Results

Population structure analysis

Principal component analysis (PCA) was performed to evaluate population stratification within the Gujarati cohort. The PCA plot demonstrated tight clustering of individuals with minimal population substructure, indicating substantial genetic homogeneity among the sampled individuals (Supplementary Fig.  1 ). These findings suggest that population stratification is unlikely to substantially confound the PRS comparisons performed in this study.

Comparison of genetic susceptibility towards autoimmune diseases between Diabetes patients and non-Diabetic controls

Among the 11 autoimmune diseases analysed, statistically significant differences in polygenic risk scores were observed only for systemic lupus erythematosus (SLE) and rheumatoid arthritis (RA) (Table  3 ). Diabetic individuals showed a significantly higher PRS for SLE compared to non-diabetic individuals (Wilcoxon Rank Sum test, W  = 21762, Effect Size = -0.00095, p  = 0.0083) (Fig.  1 ), indicating increased genetic susceptibility to lupus in the diabetic group. In contrast, non-diabetic individuals exhibited a significantly higher PRS for RA than diabetic individuals (Wilcoxon Rank Sum test, W  = 15891, Effect Size  = 0.00091, p  = 0.0276) (Fig.  2 ), suggesting a greater genetic predisposition to rheumatoid arthritis among non-diabetics. No significant differences in PRS distributions were detected for the remaining autoimmune diseases analysed ( p  > 0.05).

Table 3 Summary of polygenic risk score (PRS) comparisons for autoimmune diseases between diabetic and non-diabetic Gujarati individuals Disease Number of SNPs utilized in PRS calculation Test performed on the PRS scores of Diabetic and Non-Diabetic individuals W
p
Interpretation Autoimmune Hepatitis 5868 Wilcoxon rank sum test 20,862 0.05784 Non-significant Autoimmune Hyperthyroidism 5286 Wilcoxon rank sum test 17,351 0.3108 Non-significant Hyperthyroidism 5286 Wilcoxon rank sum test 16,685 0.1215 Non-significant ILD related to Systemic Autoimmune Disorders 5286 Wilcoxon rank sum test 16,256 0.05719 Non-significant Myasthenia Gravis 6089 Wilcoxon rank sum test 16,676 0.09414 Non-significant Autoimmune Haemolytic Anaemias 5286 Wilcoxon rank sum test 20,525 0.1056 Non-significant Psoriatic Arthritis 4433 Wilcoxon rank sum test 19,706 0.3457 Non-significant Rheumatoid Arthritis 3933 Wilcoxon rank sum test 15,891 0.02757 Significant Severe Autoimmune Type 2 Diabetes 3224 Wilcoxon rank sum test 16,030 0.03677 Significant Systemic Lupus Erythematosus 2038 Wilcoxon rank sum test 21,762 0.008327 Significant Type 1 Diabetes 5863 Wilcoxon rank sum test 17,444 0.3562 Non-significant

Summary of polygenic risk score (PRS) comparisons for autoimmune diseases between diabetic and non-diabetic Gujarati individuals

Distribution of polygenic risk scores for systemic lupus erythematosus (SLE). Violin plot showing the distribution of polygenic risk scores (PRS) for SLE in diabetic and non-diabetic individuals. The plot illustrates differences in the central tendency and spread of PRS between the two groups, with diabetic individuals exhibiting a higher median PRS, indicating increased genetic susceptibility to SLE

Distribution of polygenic risk scores for rheumatoid arthritis (RA). Violin plot depicting the distribution of polygenic risk scores (PRS) for RA in diabetic and non-diabetic individuals. Non-diabetic individuals show a higher median PRS and broader distribution, suggesting a greater genetic predisposition to RA compared to diabetic individuals

In addition to RA and SLE, the PRS analysis also identified a significant difference for severe autoimmune type 2 diabetes between diabetic and non-diabetic individuals (Wilcoxon rank-sum test, p =  0.03677). This observation is notable because severe autoimmune forms of diabetes represent conditions characterised by overlapping metabolic and immune dysregulation.

Pathway enrichment analysis

Significant SNPs selected for pathway enrichment analysis were annotated using the SNPnexus web-based server [ 30 ], which performs positional SNP-to-gene mapping based on nearby annotated genes and genomic coordinates. The resulting non-redundant gene list was used for downstream functional enrichment and pathway analysis. Because positional annotation may not fully capture distal regulatory interactions, the identified pathways were interpreted as functional associations rather than definitive mechanistic relationships. Pathway enrichment analysis revealed significant enrichment of immune- and inflammation-related pathways for both SLE (Fig. 3 a) and RA (Fig. 3 b). Prominently enriched pathways included antigen processing and presentation, allograft rejection, Th1 and Th2 cell differentiation, Th17 cell differentiation, intestinal immune network for IgA production, and hematopoietic cell lineage. Disease-associated pathways such as rheumatoid arthritis, systemic lupus erythematosus, autoimmune thyroid disease, and type 1 diabetes mellitus were also observed, reflecting shared immune-related genetic architecture.

( a ) KEGG pathway enrichment analysis for systemic lupus erythematosus (SLE) and ( b ) KEGG pathway enrichment analysis for rheumatoid arthritis (RA). KEGG pathway enrichment analysis of genes derived from SLE–associated SNPs reveals significant enrichment of immune regulatory and inflammatory pathways linked to diabetes-associated immune dysregulation. These pathways may underlie the increased genetic susceptibility to SLE observed among diabetic individuals, reflecting shared immunometabolic mechanisms between diabetes and lupus. KEGG pathway enrichment analysis of genes mapped from RA–associated SNPs highlights immune and inflammatory pathways relevant to genetic susceptibility differences between diabetic and non-diabetic individuals. Enriched pathways include antigen processing and presentation, T-helper cell differentiation, and autoimmune disease signalling, suggesting diabetes-related modulation of immune pathways that may influence RA genetic risk

In addition, several infection- and viral-response-related pathways, including Epstein–Barr virus infection, herpes simplex virus infection, and viral myocarditis, were enriched. These enrichments likely reflect shared immune-regulatory and inflammatory signalling genes commonly involved in autoimmune and host immune response pathways rather than direct evidence of viral involvement. Enrichment significance was assessed using GO and KEGG analyses with an FDR threshold of 0.01.

Discussion

In the present study, we investigated genetic susceptibility to multiple autoimmune diseases in diabetic and non-diabetic individuals from an Indian cohort using polygenic risk score (PRS) analysis. Among the 11 autoimmune disorders analysed, statistically significant differences were observed only for systemic lupus erythematosus (SLE) and rheumatoid arthritis (RA), highlighting disease-specific genetic associations with diabetic status. These findings underscore the complexity of immune–metabolic interactions and suggest that diabetes differentially modulates genetic risk across autoimmune conditions rather than exerting a uniform effect.

Diabetic individuals exhibited a higher median PRS for systemic lupus erythematosus, indicating an increased burden of lupus-associated risk alleles in this group. This observation suggests enhanced genetic susceptibility to SLE among individuals with diabetes and points toward shared immunometabolic mechanisms underlying both conditions. Diabetes is characterised by chronic low-grade inflammation, altered cytokine profiles, and immune dysregulation, which may interact with lupus-associated genetic variants to amplify disease risk. Persistent activation of innate immune pathways, metabolic stress, and impaired immune tolerance in diabetes could facilitate the expression of SLE-related genetic susceptibility, thereby predisposing diabetic individuals to lupus-like autoimmune responses.

The pathway enrichment analysis performed in this study was intended primarily to provide functional contextualization of disease-associated genetic variants underlying the observed PRS differences between diabetic and non-diabetic individuals, rather than identify entirely novel subgroup-specific biological mechanisms. Many of the enriched pathways identified, including immune signalling and inflammatory pathways, are already well established in the pathogenesis of autoimmune diseases such as RA and SLE.

Pathway enrichment analysis further supports this immunometabolic link, revealing significant enrichment of immune-related pathways such as antigen processing and presentation, Th1 and Th2 cell differentiation, Th17 cell differentiation, and hematopoietic cell lineage pathways. Dysregulation of Th1/Th2 balance and enhanced Th17 responses have been implicated in both SLE pathogenesis and metabolic inflammation, suggesting overlapping mechanisms that may bridge diabetes and lupus susceptibility. Additionally, enrichment of viral infection–related pathways, including Epstein–Barr virus and herpes simplex virus infection, aligns with existing evidence implicating viral triggers in SLE onset, potentially acting in concert with diabetes-associated immune perturbations to exacerbate autoimmune risk [ 31 – 34 ].

In contrast, non-diabetic individuals demonstrated a higher median PRS and broader distribution for rheumatoid arthritis, indicating a greater genetic predisposition to RA in the absence of diabetes. This inverse pattern suggests that diabetic individuals may carry a relatively lower genetic risk for RA, potentially reflecting disease-specific protective or antagonistic genetic mechanisms. This observation is consistent with previous genetic evidence suggesting that type 2 diabetes mellitus may act as an inverse risk factor for the development of rheumatoid arthritis [ 35 ]. Diabetes-associated genetic variants or metabolic adaptations may modulate immune pathways in a manner that reduces RA susceptibility, possibly through altered cytokine signalling, immune cell polarisation, or metabolic constraints on inflammatory responses.

The contrasting PRS patterns observed for SLE and RA emphasise that the relationship between diabetes and autoimmune diseases is highly disease-specific. While shared immunometabolic dysfunction appears to enhance genetic susceptibility to SLE, diabetes may exert a protective or modulatory effect on RA risk. While these findings highlight the heterogeneity of autoimmune pathogenesis and reinforce the concept that immune–metabolic crosstalk plays a central role in shaping disease-specific genetic risk profiles, these findings should be interpreted as supportive functional annotations of established autoimmune disease biology within the context of differential polygenic susceptibility, rather than definitive evidence of diabetes-specific autoimmune mechanisms.

An interesting observation in the present study was the significant enrichment of PRS not only for classical autoimmune disorders such as RA and SLE, but also for severe autoimmune type 2 diabetes. This provides an additional proof-of-concept support for the hypothesis that immune-regulatory and inflammatory genetic mechanisms may partially overlap between type 2 diabetes and autoimmune disorders. While T2D has traditionally been considered a metabolic disease, accumulating evidence suggests that chronic inflammation, cytokine dysregulation, and immune-mediated pathways contribute substantially to its pathogenesis. The observed PRS differences therefore appear consistent with emerging models of shared immune-metabolic genetic architecture linking autoimmune conditions and diabetic phenotypes. Rather than suggesting that classical T2D itself is an autoimmune disorder, our findings reflect shared polygenic susceptibility involving chronic inflammatory and immune-regulatory pathways. Nevertheless, these findings should be interpreted cautiously given the exploratory nature of the study and the absence of detailed autoimmune clinical phenotyping within the analysed cohort.

The findings of the present study are broadly consistent with emerging evidence suggesting shared genetic architecture between autoimmune, inflammatory, and metabolic disorders. Previous large-scale genomic studies have demonstrated substantial overlap in immune-regulatory and inflammatory pathways across autoimmune diseases, supporting the concept of shared polygenic susceptibility among clinically distinct disorders [ 36 ]. In parallel, recent PRS- and GWAS-based investigations have also identified shared genetic contributions between type 2 diabetes and other complex disorders, including cardiovascular and inflammatory conditions, indicating that metabolic diseases may partially overlap with broader immune and inflammatory genetic networks [ 37 ]. The significant PRS differences observed for rheumatoid arthritis, systemic lupus erythematosus, and severe autoimmune type 2 diabetes in the present study are therefore consistent with the growing recognition of interconnected immune-metabolic genetic mechanisms underlying complex disease susceptibility.

While the present study identified significant differences in polygenic susceptibility to certain autoimmune diseases between diabetic and non-diabetic individuals, several limitations associated with cross-ancestry PRS application should be considered while interpreting these findings. The GWAS summary statistics used for PRS construction were predominantly derived from European and East Asian populations because large-scale autoimmune GWAS datasets from Indian populations remain limited. It is well established that the predictive performance and transferability of PRS can vary substantially across populations due to differences in linkage disequilibrium (LD) structure, allele frequencies, genetic architecture, demographic history, and population-specific environmental interactions. Consequently, PRS derived from non-Indian populations may not fully capture the underlying genetic risk architecture of autoimmune diseases in Gujarati or broader Indian populations.

In the present study, principal component analysis demonstrated relatively tight clustering and genetic homogeneity within the Gujarati cohort, suggesting that internal population stratification was unlikely to substantially confound the observed group comparisons. Nevertheless, the broader challenge of cross-population PRS portability remains relevant, particularly when applying effect sizes estimated from ancestrally distinct GWAS datasets to South Asian populations.

Interestingly, the significant associations observed for rheumatoid arthritis (RA) and systemic lupus erythematosus (SLE) were derived from East Asian GWAS datasets, which may reflect comparatively improved PRS transferability between Asian populations relative to European-derived datasets. However, these findings should still be interpreted cautiously until validated in larger ancestry-matched cohorts.

Also, while statistically significant differences in PRS distributions were observed for certain autoimmune diseases, these findings should be interpreted in the context of multiple phenotype-level comparisons performed in this study. The evaluated autoimmune traits were selected based on prior biological relevance and known immune-metabolic relationships rather than through an agnostic genome-wide screening approach. Consequently, the analyses were treated as exploratory and hypothesis-driven, with nominal p-values interpreted cautiously. Further validation in larger independent cohorts and ancestry-matched populations will be necessary to confirm the robustness and reproducibility of the observed associations. Notably, environmental factors, clinical heterogeneity, and gene–environment interactions were also not explicitly modelled and may further influence autoimmune susceptibility.

Another important limitation of the present study is the absence of detailed clinical information regarding autoimmune disease diagnoses within the analysed cohort. The available dataset classified individuals based on diabetic status only, and information regarding confirmed autoimmune disease phenotypes was not available in the anonymised metadata. Consequently, the observed PRS differences should not be interpreted as evidence of clinically manifested autoimmune disorders among the study participants, but rather as differences in underlying genetic susceptibility to autoimmune diseases between diabetic and non-diabetic individuals. Future studies integrating detailed clinical phenotyping and longitudinal health records will be important to determine whether the observed polygenic susceptibility differences translate into altered autoimmune disease risk in diabetic populations.

Overall, this study provides novel insights into the complex genetic interplay between diabetes and autoimmune diseases in an Indian population. By integrating polygenic risk analysis with pathway-level interpretation, our findings underscore the importance of considering immunometabolic context when evaluating autoimmune disease susceptibility. Future studies involving larger ancestry-matched Indian cohorts with comprehensive clinical phenotyping and longitudinal follow-up will be important for validating the PRS associations identified in the present study. Integration of multi-omics approaches, including transcriptomics, epigenomics, proteomics, and metabolomics, may further improve understanding of the shared immune-metabolic mechanisms underlying autoimmune susceptibility in diabetic individuals. In addition, functional characterisation of prioritised loci and incorporation of population-specific GWAS datasets may help refine PRS portability, predictive accuracy, and translational applicability in South Asian populations.

Conclusions

Overall, our study suggests differential autoimmune genetic susceptibility profiles for selected autoimmune diseases between diabetic and non-diabetic individuals within an Indian population. A higher genetic predisposition to rheumatoid arthritis was observed among non-diabetic individuals, whereas diabetic individuals exhibited increased genetic susceptibility to systemic lupus erythematosus, suggesting a disease-specific genetic interplay between diabetes and autoimmunity, particularly in the context of lupus. These findings underscore the heterogeneous nature of autoimmune risk in relation to diabetes and emphasise that immunogenetic associations vary across autoimmune conditions.

Supplementary Information

Below is the link to the electronic supplementary material.

Supplementary Material 1

Acknowledgements

We thank Gujarat Biotechnology Research Centre (GBRC) for kindly sharing the data with us.

Abbreviations

Type 2 Diabetes

Interstitial Lung Disease

Systemic Lupus Erythematosus

Rheumatoid Arthritis

Multiple Sclerosis

Single Nucleotide Polymorphisms

Single Nucleotide Variants

Genome Wide Association Studies

Polygenic Risk Score

Gujarat Biotechnology Research Centre

Variant Call Format

Gene Ontology

Kyoto Encyclopedia of Genes and Genomes

False Discovery Rate

Author contributions

SRV: Methodology, data curation, formal analysis, and writing – first draft.AAI: Methodology, data curation, formal analysis, and writing – review and editing.RD: Conceptualization, investigation, methodology, supervision and writing – review and editing.

Funding

The Article Processing Charge (APC) was covered by Yenepoya (Deemed to be University), Mangalore, India.

Data availability

The anonymised data is available with the corresponding author upon request.

Declarations

Ethics approval and consent to participate

Ethical clearance was not required for this current study. Genomic data and associated demographic and clinical information were obtained from the Gujarat Biotechnology Research Centre (GBRC), Gujarat (Pandit et al. 2022). The aforesaid study was approved by the Institutional Ethical Committee of GBRC, Gandhinagar, B. J. Medical College and Civil Hospital, Ahmedabad, reference No. EC/Approval/38/2020 and GMERS Medical College, Gandhinagar, reference No. GMERS/MCG/IEC/06/2020. Yenepoya (Deemed to be University) was not involved in sample collection and obtaining informed consent. No individuals were sampled by the authors themselves. The initial dataset was collected following the pertinent ethical standards and guidelines of the participating institutions. All participants provided written informed consent prior to their inclusion in the analysis. In case of minors in the age range of 15–18 years, written informed consent was obtained in advance from the parents or legal guardians of the minors before the study. The use of the dataset was authorised by the GBRC for research purposes, and the data analysis was performed based on scientifically anonymised data. Our study adhered to the Declaration of Helsinki.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Footnotes

Publisher’s note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

R. V. Shravya and A. I. Ananthakrishnan contributed equally to this work.

References

Associated Data

Supplementary Materials

Supplementary Material 1

Data Availability Statement

The anonymised data is available with the corresponding author upon request.