Next Article in Journal / Special Issue
Formulation of Topical Antioxidant Creams with Hydroxycitrate or Aglianico Del Vulture Red Wine Extract for the In Vitro Prevention of Blue Light-Induced Oxidative Stress
Previous Article in Journal
The Science and Evolutionary Perspective of Beautiful Skin
Previous Article in Special Issue
Dermoscopy of Facial Dermatoses: An Updated Review
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Single-Locus, Interaction, and Functional Pathway Analyses of Acne Severity in a 60-SNP Panel

by
Valentina Russo
1,
Laura Vila-Vecilla
1,
Albert Sanchez Guerrero
1,
Laura Gascón Madrigal
1,
Caroline Brandão Chiovatto
2 and
Gustavo Torres de Souza
1,3,*
1
Fagron Genomics, 08226 Barcelona, Spain
2
Fagron Genomics Brasil, São Paulo 04571-010, SP, Brazil
3
Human Genome and Stem Cell Research Center, São Paulo University, São Paulo 05508-220, SP, Brazil
*
Author to whom correspondence should be addressed.
Cosmetics 2025, 12(5), 217; https://doi.org/10.3390/cosmetics12050217
Submission received: 20 August 2025 / Revised: 23 September 2025 / Accepted: 30 September 2025 / Published: 2 October 2025
(This article belongs to the Special Issue Feature Papers in Cosmetics in 2025)

Abstract

Acne vulgaris is a multifactorial disease with high heritability, but the genetic determinants of severity remain incompletely defined. This study evaluated 650 individuals genotyped with a 60-single-nucleotide polymorphism (SNP) panel covering immune, lipid, endocrine, and barrier pathways. Acne severity was graded as 1 (n = 193), 2–3 (n = 383), or 4 (n = 74). Single-SNP analysis highlighted associations in loci such as LHCGR (rs13405728), TGF-β2 (rs1159268), FST (rs38055), WNT10A (rs74333950), PIK3R1 (rs10515088), and THADA (rs13429458) and barrier-related variants (FLG, FLG-AS1). Epistasis analysis of 44 quality-controlled SNPs revealed 190 significant interactions (false discovery rate, FDR ≤ 0.10), with TLR4 as the main hub (degree = 22), bridging immune (IL10, TNF), lipid (PNPLA3, APOE), and barrier (FLG-AS1, OVOL1) genes. Polygenic risk scoring (PRS) showed a monotonic increase across severity grades, with Grade 4 displaying higher median scores (0.319) compared to Grade 1 (−0.129) and Grades 2–3 (0.034). Discrimination was modest but consistent (AUC: 0.661 for Grade 4 vs. 1; 0.662 vs. 2–3; 0.679 vs. all others). These results support a framework where microbial sensing, lipid metabolism, and barrier function converge to drive severe acne, underscoring the potential of genetic profiling for risk stratification and precision therapy.

1. Introduction

Acne vulgaris is one of the most common chronic inflammatory skin disorders, affecting the pilosebaceous unit and ranking among the top causes of dermatological morbidity worldwide [1,2]. Recent large-scale epidemiological studies estimate a global prevalence of approximately 20%, with peak rates approaching 28% in late adolescence and early adulthood [1,2,3,4]. Although often perceived as a condition of youth, acne can persist or emerge in adulthood. Epidemiological data show a steady increase in prevalence and disability-adjusted life years across all age groups over the past three decades. The disease exhibits marked sex- and age-related differences. Adolescent males are more likely to present with severe forms, whereas adult acne disproportionately affects females. Across diverse populations, acne is a significant public health issue. It is associated with psychosocial burden, reduced quality of life, and, in severe or untreated cases, lasting scarring [5,6,7,8].
Pathophysiologically, acne is multifactorial, arising from the interplay of increased sebum production, abnormal follicular keratinisation, colonisation by Cutibacterium acnes, and sustained inflammatory responses [9,10,11]. The clinical spectrum ranges from comedonal to severe inflammatory and nodulocystic lesions, with distribution typically on the face, chest, and back. Inflammatory lesions, especially when recurrent or inadequately managed, can lead to permanent textural changes in the skin. Meta-analytic data indicate that nearly half of individuals with acne develop clinically significant scarring, with risk increasing alongside disease severity, male sex, and positive family history. These findings underscore the need for early diagnosis and appropriate intervention to control active disease and prevent long-term complications [3,12,13,14].
Genetics contributes substantially to acne. Genome-wide association studies (GWASs) and candidate-gene studies implicate immune, endocrine, lipid, and barrier pathways in both susceptibility and severity (e.g., CDC7, SLC7A1, TNF, CYP17A1, FST, TLR4) [15,16,17,18,19,20,21,22]. Functional genomic approaches add further insight. Proteome-wide Mendelian randomisation associated FASN and TIMP4 with acne risk, while metabolomic studies suggested protective effects of metabolites such as stearoylcarnitine [17]. Together, these findings support a model in which lipid metabolism, immune modulation, keratinocyte activity, and extracellular matrix remodelling converge. This framework informs biomarker development, risk stratification, and pharmacogenetically guided interventions [16,17,18,19,20,21,22]. Acne pharmacotherapy targets the four central pathogenic processes: follicular hyperkeratinisation, proliferation of C. acnes, inflammation, and excessive sebum production [23,24]. Topical retinoids address hyperkeratinisation and dampen inflammation, while tetracyclines provide antimicrobial and host-directed anti-inflammatory effects; detailed mechanisms exceed the scope of the introduction [25,26,27,28,29,30,31]. Isotretinoin is the only therapy that modifies all four pathogenic pathways, with profound effects on sebaceous activity and inflammation; detailed mechanisms are summarised elsewhere [32,33].
Acne severity grading systems combine lesion morphology and anatomical distribution. Some scales also include secondary features such as scarring or post-inflammatory hyperpigmentation to guide diagnosis, monitoring, and therapy. Classical frameworks, including the European Dermatology Forum four-level scale and its adoption in Global Alliance consensus, stratify disease from comedonal acne (Grade 1) through papulopustular forms (Grades 2–3) to severe nodular or conglobate presentations (Grade 4) [34,35]. This clinical progression reflects underlying pathophysiology: follicular hyperkeratinisation in early stages; sustained C. acnes colonisation with innate immune activation in Grades 2–3; and, in advanced cases (Grade 3 nodular and Grade 4), granulomatous inflammation with fibrosis and scarring [36,37,38,39]. Standardised tools (e.g., IGA, GAGS, CASS, GEA) provide reproducible scoring and capture the shift from superficial to deeper inflammation. We used a four-level framework (Grade 1, Grades 2–3, Grade 4) aligned with established European/Global Alliance conventions [35,40,41].
Human genomics implicates inflammatory, sebaceous, keratinisation, and immune pathways in acne; pharmacogenomics further links genetic variation to treatment response and safety, supporting more rational, targeted care [16,17,21,42,43,44,45,46,47,48]. The ability to predict acne severity and treatment outcomes from genetic profiles could shift management toward a precision medicine model, where therapy is tailored to the individual’s molecular and clinical phenotype from the outset [49,50]. In this context, polygenic risk scores (PRSs) and multi-locus models, as explored in recent large-scale studies, provide a framework for stratifying patients according to their predicted disease course. For example, individuals with genetic signatures linked to more severe, inflammatory phenotypes could be considered earlier for systemic agents or combination regimens, while those with lower risk might benefit from optimized topical maintenance strategies. Such an approach could reduce time to effective disease control, minimise exposure to ineffective treatments, and mitigate long-term sequelae such as scarring [48,49]. The present study builds on this rationale by evaluating single-locus, interaction, and pathway-level genetic associations with acne severity in a predefined 60-SNP panel, aiming to provide evidence that can inform both mechanistic understanding and the future personalization of therapeutic decisions.

2. Materials and Methods

2.1. Database and Ethics Committee

This study used fully anonymised clinical and genetic data obtained from an internal database generated as part of routine industry activity, with all direct and indirect identifiers removed to eliminate any possibility of re-identification. No patient contact was involved, and no new data collection was undertaken for the purposes of this analysis. Therefore, an ethics-protocol exemption was granted. The use of anonymised data complies with the principles of the Declaration of Helsinki (Paragraphs 23–25 and 32) and the General Data Protection Regulation (GDPR, Recital 26), which excludes truly anonymised data from the scope of personal data regulation, ensuring adherence to EU and international standards for retrospective research using anonymised datasets.

2.2. Data Structure and SNPs

The dataset consisted of 650 fully anonymised clinical and genetic records from individuals tested with a 60-SNP pharmacogenetic acne panel, originally generated as part of routine clinical practice and stored in a secure internal database. All direct and indirect identifiers were removed before analysis, in line with the ethical and data protection principles described in Section 2.1. Clinical data included acne severity graded by the attending healthcare professional at the time of testing, using four standard categories: Grade 1, Grade 2, Grade 3, and Grade 4. For analysis purposes, Grades 2 and 3 were combined into a single category (Grade 2–3) to ensure adequate group sizes and reflect comparable clinical severity. Genetic data comprised the full set of 60 preselected single-nucleotide polymorphisms (SNPs) covering genes implicated in inflammatory response, sebaceous gland function, follicular keratinisation, androgen signalling, and lipid metabolism. These SNPs were derived from routine genotyping and were not generated specifically for this research. The 60 SNPs were selected based on prior evidence from the literature and internal studies linking them to pathways involved in acne pathophysiology, namely, inflammation, sebaceous activity, follicular keratinisation, androgen signalling, and lipid metabolism. Some SNPs had also been previously investigated in relation to pharmacogenetic responses to acne therapies. These pharmacogenetic associations were validated in earlier work as tentative correlations with treatment outcomes in clinical databases. However, as the present study was designed to focus on the mechanistic basis of acne severity, the pharmacogenetic dimension was not included in the analyses described here. The full list of SNPs and their corresponding gene annotations is presented in Table 1.

2.3. Bioinformatics and Statistical Treatment

All statistical and bioinformatic analyses were conducted in R within RStudio (Version 2025.05.1 + 513), using established packages for genetic association testing, data manipulation, and visualisation. We performed data preprocessing, quality control, and allele-frequency calculations with tidyverse and readxl. This ensured consistent handling of genotype encodings and missing data. Single-SNP association testing and multivariate models were implemented using the SNPassoc package, which allows for flexible specification of genetic models (codominant, dominant, recessive, overdominant, log-additive) and covariate adjustment within logistic regression frameworks [51]. We conducted SNP–SNP interaction analyses in R with custom epistasis scripts. We then performed functional mapping and enrichment in Cytoscape (Version 3.10.3) using ClueGO and CluePedia to identify over-represented biological processes, molecular functions, and pathways. Polygenic risk scores were calculated from significant SNPs using weighted allele counts based on regression coefficients, with performance assessed by receiver operating characteristic (ROC) curve analysis and area under the curve (AUC) estimation. All analyses adhered to recommended practices for retrospective genotype–phenotype association studies. Reproducible code and parameter settings were documented to ensure replicability. Although SNPassoc supports covariate-adjusted models, in this study, we deliberately omitted clinical covariates (sex, age, BMI, treatment exposure/history) to isolate the genetic contribution to severity within a retrospective, routinely genotyped cohort. This design choice ensures that all reported effects and the PRS reflect genetics alone, without conflation by heterogeneous or incompletely ascertained clinical variables; combined clinical–genetic models are planned for future validation work. Analysis scripts used in this study are available from the corresponding author upon reasonable request; summary statistics and PRS weights are provided in the Supplementary Materials (S1).
The initial dataset comprised a 60-SNP candidate panel with the phenotype Acnegrade (1 = grade 1; 2 = grades 2–3; 4 = grade 4), stored in the internal tables Acnecomplete/AcneSNP. Based on a priori biological suitability and availability for polygenic risk score (PRS) construction, a predefined whitelist of 44 SNPs was established for PRS and SNP–SNP interaction analyses. For these 44 SNPs, quality control metrics included call rate, minor allele frequency (MAF), and Hardy–Weinberg equilibrium exact p-value, calculated using the SNPassoc package in R. Given the sample size constraints, a relaxed QC approach was adopted to flag only clear data quality issues rather than applying stringent thresholds that could reduce statistical power. Flagged SNPs were excluded from PRS and interaction analyses but were retained for single-SNP association testing, where their potential inclusion in interpretation was further assessed according to statistical significance and functional ontology.

2.3.1. Single SNP Association

Single-locus association analyses were conducted to evaluate the relationship between each SNP in the 60-variant panel and acne severity using the SNPassoc package in R (Version 3.10.3) [51]. Acne severity was analysed both as an ordinal variable (Acnegrade: 1 = grade 1; 2 = grades 2–3; 4 = grade 4) via proportional odds logistic regression. Genotypes were coded according to reference and alternative alleles, with missing data handled under the package’s default procedures. For each SNP, association was tested under codominant, dominant, recessive, overdominant, and log-additive models using the WGassociation function. The proportional odds assumption for ordinal models was assessed within the SNPassoc framework. In addition, a Bonferroni correction was applied to codominant model p-values to account for multiple testing, with the corrected α set as 0.05 divided by the number of valid SNPs tested. No multiple testing correction was applied at this stage, as the aim was to identify both strong candidates and variants of potential biological relevance for subsequent interaction, enrichment, and PRS analyses.

2.3.2. SNP–SNP Epistasis Network and Functional Enrichment Analysis

SNP–SNP interaction (epistasis) analysis was performed on the 44 preselected SNPs from the whitelist, derived from the original 60-SNP panel, in 650 individuals graded for acne severity on a three-level ordinal scale (1 = grade 1, 2 = grades 2–3, 4 = grade 4). Genotypes were recoded as allele dosages (0, 1, 2) with orientation such that the homozygote for the minor allele was coded as 2. Missing values were mean-imputed per SNP, and all dosage values were z-standardised prior to modelling. For each SNP pair, we fitted two proportional-odds logistic regression models: a base model containing both SNPs as main effects and an extended model including the SNP × SNP interaction term. Likelihood ratio tests were used to compare models, with the p-value from the interaction term retained as the measure of epistatic evidence. We also recorded the direction of the interaction coefficient, the number of genotyped individuals for the pair, and the pairwise linkage disequilibrium (LD, r2) between SNPs, calculated from the standardised dosages [52].
Multiple testing correction was applied using the Benjamini–Hochberg false discovery rate (FDR). Significant epistatic pairs were defined as those with FDR ≤ 0.10 and LD r2 ≤ 0.30, with a relaxed view considered at FDR ≤ 0.20. When no pairs met the threshold, the top-ranking pairs by p-value (default N = 25) were retained for exploratory analysis. Each selected SNP pair was mapped to its corresponding gene(s) according to the curated SNP–gene correspondence list provided for this study. For SNPs mapping to multiple genes, all gene–gene combinations were retained. Gene–gene edges were assigned a weight corresponding to the maximum −log10(FDR) from the supporting SNP pairs, together with the mean LD across those pairs and the number of supporting SNP pairs. Node degree, defined as the number of unique edges incident on a given gene, was then computed for the network analysis.
Functional enrichment analysis was performed on the resulting set of genes from the epistasis network using ClueGO and CluePedia plug-ins in Cytoscape (Version 3.10.3). Analyses focused on Gene Ontology categories, with the full set of genes mapped from the whitelist serving as the background. p-values were adjusted for multiple testing using the Benjamini–Hochberg method, and results were reported at an adjusted q-value threshold of ≤0.20, with a relaxed view at ≤0.25. Redundant Gene Ontology (GO) terms were reduced based on functional grouping, and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis was performed when annotation was available. Outputs included enrichment tables, term–gene correspondence files, and Cytoscape-ready formats for network visualisation, enabling the representation of top enriched terms, term–gene associations, and term–term similarity networks based on shared gene content [53,54]. Functional enrichment used GO Biological Process in ClueGO/CluePedia with the whitelist-mapped genes as the background. For each term, we exported both raw and adjusted p-values from the ClueGO output. Results were interpreted as suggestive given the exploratory thresholds.

2.3.3. Polygenic Risk Score Evaluation

The polygenic risk score (PRS) was developed using a prespecified whitelist of 44 candidate SNPs from the original 60-SNP panel after quality control treatment, with acne severity coded as an ordinal outcome (1 = grade 1, 2 = grades 2–3, 4 = grade 4). Genotypes were coerced to allele dosages (0/1/2) and oriented to the minor allele in the training data of each fold. Quality control was performed within folds to avoid information leakage, applying relaxed thresholds suitable for the available sample size: per-sample call rate ≥ 0.85 (fallback 0.80), per-SNP call rate ≥ 0.90 (fallback 0.85), minor allele frequency ≥0.01 (fallback 0.005), with Hardy–Weinberg equilibrium p-values recorded but not used as strict exclusion criteria. Missing values were imputed by the mean dosage per SNP, followed by z-score standardisation. Linkage disequilibrium pruning removed one SNP from any pair with r2 > 0.90, favouring higher call rate and MAF.
Model fitting used penalised ordinal regression with a cumulative logit link (package ordinalNet), applying elastic-net regularisation with mixing parameter α ∈ {1.0, 0.5, 0.25, 0.10} and tuning λ by inner cross-validation. We used nested stratified 5 × 5 cross-validation to generate out-of-fold predictions and obtain unbiased performance estimates. When the ordinal model did not converge, we fitted a multinomial elastic-net model (glmnet) and collapsed class-specific coefficid to a single PRS aligned with increasing severity. We quantified discrimination with Harrell’s c-index and binary AUCs for key contrasts, assessed overall performance with multiclass and binary Brier scores, and evaluated calibration using intercept and slope from logistic recalibration models. The final PRS was refitted on the entire dataset using identical preprocessing, and SNP weights were exported along with performance summaries and figures. All analyses were conducted in R (v4.4.1) using the packages ordinalNet, glmnet, pROC, Hmisc, HardyWeinberg, caret, ggplot2, openxlsx, and ragg.

3. Results

3.1. Overall Database Description and Quality Control Results

The study dataset comprised 650 individuals tested with the 60-SNP acne pharmacogenetic panel, including 167 males and 483 females. Acne severity at the time of testing was distributed as follows: 193 individuals with Grade 1 (mild, non-inflammatory), 383 with Grades 2–3 (moderate inflammatory), and 74 with Grade 4 (severe nodular/conglobate) acne. A stratified baseline table by severity and sex is provided in Supplementary File S1; age and ancestry data were not available in this cohort. Quality control (QC) was applied to a predefined whitelist of 44 SNPs selected a priori for polygenic risk score (PRS) and SNP–SNP interaction analyses, while all 60 SNPs were retained for single-locus association testing. QC metrics included per-sample and per-SNP call rate, minor allele frequency (MAF), and Hardy–Weinberg equilibrium (HWE) exact test p-values, computed using the SNPassoc package in R. Given the moderate sample size, relaxed thresholds were applied to flag only clear data quality issues: per-sample call rate ≥ 0.80, per-SNP call rate ≥ 0.85, and MAF ≥ 0.005. We assessed HWE for all SNPs. Variants with extreme deviations (very small p-values) were flagged and excluded only if they also failed call-rate criteria.
Following this procedure, 44 SNPs passed QC and were retained for all PRS and interaction analyses, while the remaining variants were excluded from these analyses but kept for single-locus association testing to maximise discovery potential. The retained SNPs displayed call rates between 0.972 and 1.000, MAF values ranging from 0.0346 to 0.4808, and HWE p-values from nominally non-significant to values indicating marked deviation from equilibrium. These QC-passed SNPs are summarised in Table 2, which lists the call rate, MAF, and HWE p-value for each variant according to the final relaxed thresholds applied.

3.2. Single-Locus Genotype–Phenotype Association

We analysed the 60-variant panel under five inheritance models. The additive model was prespecified as the primary analysis; codominant, dominant, recessive and overdominant models were used as sensitivity analyses. Several loci remained significant after Bonferroni correction for 60 tests (α = 8.3 × 10−4). The clearest signal was CYP3A4 rs28988604 (additive p = 3.15 × 10−8). Additional SNPs that met the Bonferroni threshold in the additive model were PIK3R1 rs10515088 (p = 3.83 × 10−7), RXR rs10918169 (p = 2.71 × 10−4), THADA rs13429458 (p = 3.97 × 10−4), TGF-β2 rs1159268 (p = 6.56 × 10−5), FST rs38055 (p = 3.59 × 10−4), and CYP2C9 rs28371686 (p = 7.44 × 10−4). Several loci were also significant only under alternate models: WNT10A rs74333950 (recessive p = 7.58 × 10−6; codominant p = 3.28 × 10−5), LHCGR rs13405728 (recessive p = 2.59 × 10−5; codominant p = 1.43 × 10−4), CYP2C9 rs1057910 (recessive p = 1.58 × 10−4; codominant p = 3.20 × 10−4), and HNF1A-AS1 rs2650000 (recessive p = 4.98 × 10−4).
Other variants showed nominal or model-specific associations but did not survive multiple-testing correction, including loci in HNF1A-AS1 (other models), C11orf30/LRRC32, TLR4, SLCO1B1, PNPLA3, SOAT1, SERPINB7, CYP19A1, FLG-AS1, and FLG. Classical inflammatory cytokine genes (IL1B, IL10, TNF) did not show significant single-locus effects in this dataset. The distribution of signals across models is depicted in Figure 1, where the SNPs listed above can be seen exceeding the Bonferroni line. Taken together, the single-variant results highlight biologically coherent axes that align with the rest of the study: retinoid metabolism (CYP3A4, CYP2C9), nuclear receptor and PI3K signalling (RXR, PIK3R1), endocrine regulation (LHCGR, THADA), tissue remodelling (TGF-β2, FST), and follicular development (WNT10A). All single-locus p-values under each inheritance model, including Bonferroni-adjusted results, are provided in Supplementary File S1.

3.3. SNP–SNP Epistasis Network and Enriched Functional Pathways

All 44 prespecified SNPs passed the relaxed QC thresholds (per-SNP call rate ≥ 0.85, MAF ≥ 0.01), with observed call rates between 0.97 and 1.00 and MAF values from 0.03 to 0.48. Hardy–Weinberg p-values were logged for reference (including some extreme deviations), but no SNPs were excluded at this stage. Interaction testing was performed for all 946 unique SNP pairs using proportional-odds logistic regression with both main effects retained. Multiplicity control via Benjamini–Hochberg FDR identified 190 SNP pairs meeting FDR ≤ 0.10, with p-values from 1.10 × 10−6 to ~2.01 × 10−2. A consolidated table of the 190 significant SNP–SNP pairs (FDR ≤ 0.10) is provided in Supplementary File S1. The top-ranked interactions generally occurred in low linkage disequilibrium (most r2 < 0.15). High-ranking examples included SERPINB7– rs873549 (p = 1.10 × 10−6, r2 = 0.0063), FLG-AS1–TLR4 (p = 1.25 × 10−6, r2 = 0.0019), THADA–SLCO1B1 (p = 1.38 × 10−6, r2 = 0.270), IL10–TLR4 (p = 4.39 × 10−6, r2 = 0.0034), and FSHR–HLA-DRB1 (p = 4.94 × 10−6, r2 = 0.026). Additional strong associations were observed for lipid/inflammation-related pairs such as PNPLA3–RETN and MYC–PNPLA3 (see Figure 2).
To prioritise the robustness of the epistasis model, we also examined a stricter subset meeting FDR ≤ 0.05. Several top interactions satisfied this criterion, including FLG-AS1–TLR4 (p = 1.25 × 10−6; r2 = 0.0019), IL10–TLR4 (p = 4.39 × 10−6; r2 = 0.0034), SLCO1B1–THADA (p = 1.38 × 10−6; r2 = 0.270), FSHR–HLA-DRB1 (p = 4.94 × 10−6; r2 = 0.026), and SERPINB7–intergenic rs873549 (p = 1.10 × 10−6; r2 = 0.0063). These pairs are biologically coherent with the rest of the study, mapping onto innate microbial sensing centred on TLR4 (with cytokine partner IL10 and barrier partner FLG-AS1), lipid–endocrine crosstalk (SLCO1B1–THADA), and immune–barrier signalling (FSHR–HLA-DRB1, SERPINB7–rs873549), while occurring at generally low inter-SNP linkage disequilibrium (LD), which reduces the likelihood of artefacts from correlated genotypes.
Collapsing SNP-level results to genes yielded a connected network of 39 nodes spanning innate immunity/inflammation (TLR4, IL10, IL1B, TGFB2, TNF), lipid handling (PNPLA3, TM6SF2, APOE, ABCG8, SLCO1B1, SOAT1), endocrine/reproductive signalling (FSHR, LHCGR, CYP17A1, CYP19A1), epidermal barrier–keratinisation (FLG-AS1, OVOL1, SERPINB7), and transcriptional regulation (MYC, IRF4, MYEF2, MTA3, HNF1A-AS1). TLR4 emerged as the most connected hub (degree = 22), followed by SLCO1B1 (20), PNPLA3, THADA, and HNF1A-AS1 (16 each), with RETN and MYEF2 close behind (15 each). The strongest gene–gene edges by statistical weight (−log10 p) included SERPINB7– rs873549 (5.96), FLG-AS1–TLR4 (5.90), SLCO1B1–THADA (5.86), IL10–TLR4 (5.36), and PNPLA3–RETN (5.14). Figure 3 shows the complete interaction network with top hubs highlighted in yellow, illustrating the central role of TLR4 in bridging inflammation-related and metabolic/epithelial pathways. Figure 4 presents a bar graph showing the top 12 Gene Ontology biological processes identified by ClueGO after filtering out broad reproductive terms. Enrichment significance is expressed as −log10(p), highlighting pathways most relevant to acne pathophysiology. The predominant categories include lipid localisation, epithelial cell proliferation, inflammatory response, cytokine regulation, and immune activation, underscoring the convergence of barrier, metabolic, and immune mechanisms in severe disease. Analysis parameters: GO Biological Process in ClueGO/CluePedia; background = whitelist-mapped genes; multiple testing = BH FDR. Displayed values are −log10(p) for clarity, whereas BH-adjusted q values were used for statistical interpretation. Enrichments are reported as suggestive under the exploratory thresholds used. A summary table of the top enriched Gene Ontology (GO) biological processes, including raw P and BH-adjusted q values, is provided in Supplementary File S1.

3.4. Polygenic Risk Score

The standardised PRS distribution demonstrated a clear upward shift with increasing acne severity (Figure 5). Median values were lowest for Grade 1, intermediate for Grades 2–3, and highest for Grade 4, with the Grade 4 group showing a visibly wider spread. This monotonic trend was accompanied by a moderate separation between severe and non-severe groups, reflected in the overall ordinal discrimination (c-index = 0.574) and macro AUC = 0.618 across binary clinical contrasts. When focusing on the clinically relevant comparison between severe cases and the rest, the PRS yielded an AUC of 0.679 for Grade 4 versus Grades 1/2–3. This level of discrimination, while modest, was consistent across all tested contrasts, suggesting that the genetic component captured by the PRS provides measurable but not exclusive predictive information. Because the models were intentionally unadjusted for clinical covariates (sex, age, BMI, treatment), these discrimination metrics quantify the genetic signal in isolation. As such, they likely represent a lower bound on attainable performance; incremental gains are expected when genetics is integrated with clinical and demographic predictors in external validation.
In the final penalised model fit on the full dataset, 8 of the 44 SNPs received non-zero weights: rs10515088 (PIK3R1) − 0.404846; rs38055 (FST) − 0.230665; rs873549 (intergenic) + 0.159581; rs1862513 (RETN) + 0.096053; rs13405728 (LHCGR) − 0.094171; rs12478601 (THADA) − 0.062663; rs7927894 (FLG) − 0.026665; rs717620 (ABCC2) + 0.001324. All remaining SNPs had weights of 0.000 in the penalised model. Calibration from logistic analysis showed intercept −0.150 (95% CI −0.648 to 0.347) and slope 1.167 (0.599 to 1.736) for Grade ≥2 versus Grade 1, and intercept 2.843 (−0.797 to 6.483) and slope 2.455 (0.624 to 4.287) for Grade 4 versus the rest.
ROC analyses confirmed that the model performed similarly for Grade 4 versus Grade 1 (AUC = 0.661) and Grade 4 versus Grade 2–3 (AUC = 0.662), with nearly overlapping curves (Figure 6). Collapsing Grades 1 and 2 into a single non-severe group yielded the same AUC of 0.662 (Figure 7), reinforcing the stability of the discrimination irrespective of the comparison set. Risk gradient analysis revealed that the proportion of Grade 4 cases rose from 7.7% in the lowest PRS quintile to 20.3% in the highest, albeit with some fluctuation in intermediate quintiles. Kernel density plots (Figure 8) illustrate this shift in distribution, with the severe group’s density peak displaced to the right relative to lower grades. The final PRS incorporated all 44 candidate SNPs, but penalisation concentrated weight on a small subset, with the largest absolute coefficients observed for variants in PIK3R1, FST, RETN, LHCGR, THADA, and FLG. These results indicate that while the PRS alone is not sufficient for clinical stratification, it captures meaningful genetic variation associated with severity and may complement other predictors in multi-modal models.

4. Discussion

The present study examined genetic contributions to acne severity using a predefined 60-SNP panel, combining single-locus associations, interaction networks, and polygenic risk scoring. The results offer new insights into how pathways of immunity, lipid metabolism, endocrine signalling, and skin barrier biology intersect to shape clinical outcomes. Importantly, these findings go beyond isolated associations, showing how different biological domains converge in patients with more severe disease.
At the single-variant level, associations were detected in genes involved in signalling and receptor activity (RXR, LHCGR) [55,56,57], and regulators of follicular biology and tissue remodelling (FST, THADA, WNT10A, PIK3R1) [58,59,60,61]. These loci are consistent with established mechanisms of acne pathophysiology, such as androgen metabolism, retinoid signalling, and matrix turnover. Interestingly, inflammatory cytokine genes such as IL1B, IL10, and TNF did not show significant effects in isolation, indicating that their contribution may arise in combination with other pathways rather than through direct single-locus effects [62,63,64].
The interaction analysis revealed a more complex picture, with TLR4 emerging as the most connected node in the gene–gene network. Variants in TLR4 are known to modify recognition of microbial ligands from Cutibacterium acnes, and previous case–control studies have linked TLR4 polymorphisms to greater risk of severe forms, including acne conglobata [65,66,67,68]. In our network, TLR4 interacted not only with immune mediators such as IL10 and TNF but also with genes related to lipid metabolism (PNPLA3, APOE, SOAT1), epidermal differentiation (FLG-AS1, OVOL1), and endocrine axes (FSHR, LHCGR, CYP17A1). This wide connectivity suggests that TLR4 serves as a pleiotropic regulator where microbial sensing intersects with sebaceous lipid composition, inflammatory tone, and keratinocyte function. Such convergence may explain why certain variants predispose to the most aggressive phenotypes, where microbial and immune drivers amplify one another against a background of altered lipid and barrier biology.
The role of FLG, filaggrin, and its antisense regulatory locus (FLG-AS1) also deserves emphasis. Filaggrin is central to skin barrier integrity, keratin aggregation, and hydration. Loss-of-function mutations in FLG, well recognised in atopic dermatitis, have been reported to reduce the likelihood of acne, suggesting that certain barrier alterations may protect against follicular occlusion or alter microbial colonisation [69,70,71,72]. In our network, the FLG-AS1–TLR4 interaction is particularly relevant: impaired barrier function could enhance microbial antigen penetration, priming an exaggerated inflammatory response mediated by TLR4. The integration of barrier and immune pathways points to a model where genetic determinants of keratinisation not only affect lesion formation but also shape downstream inflammatory trajectories.
Beyond immunity and barrier function, the interaction network also revealed modules linked to endocrine and metabolic regulation. Genes such as FSHR, LHCGR, CYP17A1, and CYP19A1 connect reproductive signalling with inflammatory and lipid pathways [16,17], reflecting the well-established influence of androgens on sebaceous gland activity. Lipid-related loci such as PNPLA3, TM6SF2, and APOE were repeatedly connected to inflammatory and transcriptional regulators, reinforcing the idea that sebum composition and lipid handling are genetically intertwined with immune activation in acne. These multi-domain connections provide a more nuanced view than single-gene approaches, capturing the interplay between sebaceous, immune, and barrier systems.
The polygenic risk score (PRS) analysis showed that aggregated genetic variation explains measurable differences in severity. The standardised PRS showed a monotonic increase across severity grades, with mean values of −0.129 in Grade 1, 0.034 in Grades 2–3, and 0.319 in Grade 4. Patients in the highest quintile had a 20.3% prevalence of severe disease compared to 7.7% in the lowest, representing a 2.6-fold difference. Although discrimination was modest (AUC ≈ 0.66–0.68), the consistency across contrasts and the visible risk gradient support the signal. Severity-related genetic information can be captured by a cumulative score. Variants with the strongest weights in the PRS overlapped with biologically relevant loci from the other analyses, including PIK3R1, FST, RETN, LHCGR, and FLG, underscoring the coherence of the genetic signal [19,55,59,60,61,73,74].
Taken together, the single-locus, interaction, and PRS findings highlight three interconnected axes central to acne severity: microbial sensing/immune signalling (anchored on TLR4 and cytokine pathways), sebaceous lipid metabolism (PNPLA3, TM6SF2, APOE), and barrier–keratinisation biology (FLG, OVOL1, SERPINB7) [74,75,76,77]. The convergence of these axes provides a plausible mechanistic framework for why certain individuals develop nodular or conglobate forms while others present with milder phenotypes. It also points toward therapeutic strategies that may target upstream regulators such as innate immune receptors or barrier function, in addition to conventional approaches focused on sebum suppression or antimicrobial therapy.
From a translational perspective, these results suggest that genetic information could be used to refine risk stratification and guide therapy selection. Patients with genetic signatures pointing to strong inflammatory or immune components could be considered for earlier systemic therapy (e.g., anti-inflammatory agents or isotretinoin), whereas barrier-weighted profiles may warrant barrier-supportive regimens alongside standard care; lipid/endocrine-weighted profiles may support earlier sebum-modulating strategies. The PRS, once externally validated, could help identify patients at greater risk of scarring forms, reducing the time to effective treatment and avoiding prolonged exposure to ineffective regimens. In this study, we used AUC primarily as an internal validation tool to confirm that the multi-locus genetic architecture aligned with the biological axes highlighted by the single-locus and interaction analyses—immunity/TLR4, lipid metabolism, and barrier pathways. Although discrimination was modest (AUC ~0.66–0.68), the consistent risk gradient across comparisons supports biological relevance at this effect size. To temper translational claims, we explicitly note that external validation and integration with clinical and demographic predictors will be necessary before any clinical use.
Despite these promising findings, several limitations must be acknowledged. The study included 650 individuals, but only 74 had Grade 4 acne. This restricted power for the most severe comparisons and limited PRS calibration. The SNP panel was predefined rather than genome-wide, so additional relevant loci may not have been captured. Finally, we deliberately excluded clinical covariates (sex, age, BMI, and treatment exposure/history) to isolate genetic effects within this retrospective dataset; while this clarifies the genetic signal, it limits immediate clinical applicability. Future work will explicitly test the incremental value of our PRS when combined with standardised clinical and demographic predictors in larger, multi-ethnic cohorts. Some SNPs deviated from Hardy–Weinberg equilibrium and were retained only for single-SNP evaluation to maximise discovery; these require careful re-examination in future studies. Importantly, the deliberate exclusion of clinical covariates such as sex, BMI, and treatment history allowed isolation of genetic effects but reduces immediate clinical applicability.
Future work should focus on validating these findings in larger, multi-ethnic cohorts and testing the incremental predictive value of the PRS when combined with demographic and clinical variables. Functional studies on TLR4 polymorphisms and their role in responses to Cutibacterium acnes, particularly in the context of acne conglobata, could clarify causal mechanisms. Similarly, investigation into how FLG variants modify follicular keratinisation and microbial colonisation could open avenues for barrier-based interventions. Integrating genetic, clinical, and molecular data will be essential to translate these findings into precision acne management.

5. Conclusions

This study shows the influence of interactions among immune, lipid, endocrine, and barrier pathways. TLR4 emerges as a central hub connecting microbial recognition, sebaceous lipid metabolism, and inflammatory signalling, while FLG and related loci highlight the contribution of keratinisation and barrier biology. Single-locus signals, epistatic interactions, and polygenic risk scores consistently converged on these axes, showing that aggregated genetic variation explains measurable differences in severity and distinguishes individuals at risk of nodular or conglobate forms. Consistent with this framework, PRS discrimination was intentionally evaluated as an internal check of mechanistic coherence; despite modest AUCs (~0.66–0.68), the stable risk gradient across contrasts supports biological plausibility. These findings provide a biologically coherent framework that links microbial sensing, barrier integrity, and sebaceous activity to clinical outcomes, supporting the potential of genetics to refine severity prediction and guide therapeutic decision-making.
While the translational potential is clear, this study has limitations: a moderate sample size with under-representation of Grade 4 cases, reliance on a predefined 60-SNP panel rather than genome-wide coverage, and no direct experimental validation of the interaction signals. The polygenic risk score showed modest discrimination (AUC ≈ 0.66–0.68) and was used primarily as an internal check of mechanistic coherence; nevertheless, the consistent risk gradient across severity groups supports biological relevance. Future work should validate these findings in larger, multi-ethnic cohorts, integrate polygenic scores with clinical and demographic covariates, and conduct functional studies of key loci such as TLR4 and FLG. These steps will be essential before genetic insights can be applied to guide precision acne management.

Supplementary Materials

The following supporting information can be downloaded at: https://www.mdpi.com/article/10.3390/cosmetics12050217/s1. Statistical data: Supplementary File S1.

Author Contributions

Conceptualization, G.T.d.S.; methodology, L.G.M.; software, A.S.G.; validation, V.R. and G.T.d.S.; formal analysis, V.R. and G.T.d.S.; investigation, G.T.d.S.; resources, L.V.-V.; data curation, G.T.d.S.; writing—original draft preparation, G.T.d.S.; writing—review and editing, C.B.C. and G.T.d.S.; visualization, G.T.d.S.; supervision, V.R.; project administration, L.V.-V. All authors have read and agreed to the published version of the manuscript.

Funding

This research received no external funding.

Institutional Review Board Statement

Ethical review and approval were waived for this study because it exclusively involved the analysis of fully anonymised clinical and genetic data obtained from an internal database created as part of routine industry activity. No identifiable information was used, no new data were collected, and no patient contact occurred. In accordance with the Declaration of Helsinki (Paragraphs 23–25 and 32) and the General Data Protection Regulation (GDPR, Recital 26), research using anonymised data that precludes the re-identification of individuals does not require additional ethical review. This framework ensures compliance with international ethical principles while eliminating risks related to patient privacy.

Informed Consent Statement

Patient consent was waived due to all patient data being deidentified.

Data Availability Statement

Data available upon request.

Conflicts of Interest

Valentina Russo, Laura Vila-Vecilla, Albert Sanchez Guerrero, Laura Gascón Madrigal, Caroline Brandão Chiovatto, and Gustavo Torres de Souza were employed by Fagron Genomics (Spain or Brazil). The authors declare that, regardless of the affiliation, the research was conducted in the absence of any commercial or financial interest that could be construed as a potential conflict of interest.

References

  1. Witkam, W.C.A.M.; Dal Belo, S.E.; Pourhamidi, S.; Raynaud, E.; Moreau, M.; Aguilar, L.; Jansen, P.W.; Nijsten, T.E.C.; Pardo, L.M. The Epidemiology of Acne Vulgaris in a Multiethnic Adolescent Population from Rotterdam, the Netherlands: A Cross-Sectional Study. J. Am. Acad. Dermatol. 2024, 90, 552–560. [Google Scholar] [CrossRef]
  2. Deng, J.; Peng, S.; Yang, F.; Wei, X.; Lu, X.; Lu, L.; Lu, Y.; Chen, Z.; Yan, W.; Huang, X. Global Pattern, Trend, and Cross-Country Health Inequality of Adult Acne Aged 25 + Years from 1990 to 2021, a Comprehensive Analysis for Global Burden of Disease and Global Dietary Database. J. Health Popul. Nutr. 2025, 44, 238. [Google Scholar] [CrossRef]
  3. Saurat, J.-H.; Halioua, B.; Baissac, C.; Cullell, N.P.; Ben Hayoun, Y.; Aroman, M.S.; Taieb, C.; Skayem, C. Epidemiology of Acne and Rosacea: A Worldwide Global Study. J. Am. Acad. Dermatol. 2024, 90, 1016–1018. [Google Scholar] [CrossRef]
  4. Guguluș, D.L.; Vâță, D.; Popescu, I.A.; Pătrașcu, A.I.; Halip, I.A.; Mocanu, M.; Solovăstru, L.G. The Epidemiology of Acne in the Current Era: Trends and Clinical Implications. Cosmetics 2025, 12, 106. [Google Scholar] [CrossRef]
  5. Leskelä, M.; Jokelainen, J.; Huilaja, L.; Sinikumpu, S.-P. Adult Acne in Middle-Age: Effects on Mental Health in General Population of the Northern Finland Birth Cohort 1966. Acta Derm. Venereol. 2024, 104, adv14733. [Google Scholar] [CrossRef] [PubMed]
  6. Dias da Rocha, M.A.; Saint Aroman, M.; Mengeaud, V.; Carballido, F.; Doat, G.; Coutinho, A.; Bagatin, E. Unveiling the Nuances of Adult Female Acne: A Comprehensive Exploration of Epidemiology, Treatment Modalities, Dermocosmetics, and the Menopausal Influence. Int. J. Womens Health 2024, 16, 663–678. [Google Scholar] [CrossRef]
  7. Chernyshov, P.V.; Zouboulis, C.C.; Tomas-Aragones, L.; Jemec, G.B.; Manolache, L.; Tzellos, T.; Sampogna, F.; Evers, A.W.M.; Dessinioti, C.; Marron, S.E.; et al. Quality of Life Measurement in Acne. Position Paper of the European Academy of Dermatology and Venereology Task Forces on Quality of Life and Patient Oriented Outcomes and Acne, Rosacea and Hidradenitis Suppurativa. J. Eur. Acad. Dermatol. Venereol. 2018, 32, 194–208. [Google Scholar] [CrossRef] [PubMed]
  8. Kansal, N.K.; Bhatia, R.; Rawat, V.S.; Sushantika; Vasisht, S.; Kumar, A. Psychological Aspects of Acne Vulgaris: A Narrative Review. Clin. Dermatol. Rev. 2025, 9, 109–112. [Google Scholar] [CrossRef]
  9. Cros, M.P.; Mir-Pedrol, J.; Toloza, L.; Knödlseder, N.; Maruotti, J.; Zouboulis, C.C.; Güell, M.; Fábrega, M.-J. New Insights into the Role of Cutibacterium Acnes-Derived Extracellular Vesicles in Inflammatory Skin Disorders. Sci. Rep. 2023, 13, 16058. [Google Scholar] [CrossRef] [PubMed]
  10. Shafiuddin, M.; Huang, W.; Prather, G.W.; Anton, J.R.; Martin, A.L.; Sillart, S.B.; Tang, J.Z.; Vittori, M.R.; Prinsen, M.J.; Ninneman, J.J.; et al. Cutibacterium Adaptation to Life on Humans Provides a Potential C Acnes Infection Biomarker. J. Investig. Dermatol. 2025, in press. [Google Scholar] [CrossRef]
  11. Mayslich, C.; Grange, P.A.; Dupin, N. Cutibacterium Acnes as an Opportunistic Pathogen: An Update of Its Virulence-Associated Factors. Microorganisms 2021, 9, 303. [Google Scholar] [CrossRef]
  12. Wang, Y.; Xiao, S.; Ren, J.; Zhang, Y. Analysis of the Epidemiological Burden of Acne Vulgaris in China Based on the Data of Global Burden of Disease 2019. Front. Med. 2022, 9, 939584. [Google Scholar] [CrossRef]
  13. Liu, L.; Xue, Y.; Chen, Y.; Chen, T.; Zhong, J.; Shao, X.; Chen, J. Prevalence and Risk Factors of Acne Scars in Patients with Acne Vulgaris. Ski. Res. Technol. 2023, 29, e13386. [Google Scholar] [CrossRef] [PubMed]
  14. Dall’Oglio, F.; Nasca, M.R.; Fiorentini, F.; Micali, G. Diet and Acne: Review of the Evidence from 2009 to 2020. Int. J. Dermatol. 2021, 60, 672–685. [Google Scholar] [CrossRef] [PubMed]
  15. Bataille, V.; Snieder, H.; MacGregor, A.J.; Sasieni, P.; Spector, T.D. The Influence of Genetics and Environmental Factors in the Pathogenesis of Acne: A Twin Study of Acne in Women. J. Investig. Dermatol. 2002, 119, 1317–1322. [Google Scholar] [CrossRef]
  16. Van Steensel, M.A.M. The Genetics of Acne. Ann. Hum. Genet. 2025, 89, 333–341. [Google Scholar] [CrossRef]
  17. Teder-Laving, M.; Kals, M.; Reigo, A.; Ehin, R.; Objärtel, T.; Vaht, M.; Nikopensius, T.; Metspalu, A.; Kingo, K. Genome-Wide Meta-Analysis Identifies Novel Loci Conferring Risk of Acne Vulgaris. Eur. J. Hum. Genet. 2024, 32, 1136–1143. [Google Scholar] [CrossRef]
  18. Ju, R.; Ying, Y.; Zhou, Q.; Cao, Y. Exploring Genetic Drug Targets in Acne Vulgaris: A Comprehensive Proteome-Wide Mendelian Randomization Study. J. Cosmet. Dermatol. 2024, 23, 4223–4229. [Google Scholar] [CrossRef] [PubMed]
  19. Zhang, H.; Zhang, Z. Genetic Variants Associated with Acne Vulgaris. Int. J. Gen. Med. 2023, 16, 3843–3856. [Google Scholar] [CrossRef]
  20. Wang, X.; Wu, Y.; Zhao, P.; Wang, X.; Wu, W.; Yang, J. The Causal Relationship between Serum Metabolites and Acne Vulgaris: A Mendelian Randomization Study. Sci. Rep. 2024, 14, 11045. [Google Scholar] [CrossRef]
  21. Heng, A.H.S.; Say, Y.-H.; Sio, Y.Y.; Ng, Y.T.; Chew, F.T. Gene Variants Associated with Acne Vulgaris Presentation and Severity: A Systematic Review and Meta-Analysis. BMC Med. Genom. 2021, 14, 103. [Google Scholar] [CrossRef]
  22. Ryguła, I.; Pikiewicz, W.; Kaminiów, K. Impact of Diet and Nutrition in Patients with Acne Vulgaris. Nutrients 2024, 16, 1476. [Google Scholar] [CrossRef]
  23. Lai, Y.; Fan, M.; Fan, X.; Chen, J.; Xiang, L.F.; Ma, Y. Progress on Multiomics Research on Acne Vulgaris: A Literature Review. J. Investig. Dermatol. 2025, 145, 2162–2169. [Google Scholar] [CrossRef]
  24. Li, Y.; Hu, X.; Dong, G.; Wang, X.; Liu, T. Acne Treatment: Research Progress and New Perspectives. Front. Med. 2024, 11, 1425675. [Google Scholar] [CrossRef]
  25. Kim, H.J.; Kim, Y.H. Exploring Acne Treatments: From Pathophysiological Mechanisms to Emerging Therapies. Int. J. Mol. Sci. 2024, 25, 5302. [Google Scholar] [CrossRef] [PubMed]
  26. Agamia, N.F.; El Mulla, K.F.; Alsayed, N.M.; Ghazala, R.M.; El Maksoud, R.E.A.; Abdelmeniem, I.M.; Talaat, I.M.; Zaki, I.I.; Sabah, R.M.; Melnik, B.C. Isotretinoin Treatment Upregulates the Expression of P53 in the Skin and Sebaceous Glands of Patients with Acne Vulgaris. Arch. Dermatol. Res. 2022, 315, 1355–1365. [Google Scholar] [CrossRef] [PubMed]
  27. Nelson, A.M.; Cong, Z.; Gilliland, K.L.; Thiboutot, D.M. TRAIL Contributes to the Apoptotic Effect of 13-Cis Retinoic Acid in Human Sebaceous Gland Cells. Br. J. Dermatol. 2011, 165, 526–533. [Google Scholar] [CrossRef]
  28. Dreno, B.; Chavda, R.; Julia, V.; Khammari, A.; Blanchet-Réthoré, S.; Krishnaswamy, J.K. Transcriptomics Analysis Indicates Trifarotene Reverses Acne-Related Gene Expression Changes. Front. Med. 2021, 8, 745822. [Google Scholar] [CrossRef]
  29. Cosio, T.; Di Prete, M.; Gaziano, R.; Lanna, C.; Orlandi, A.; Di Francesco, P.; Bianchi, L.; Campione, E. Trifarotene: A Current Review and Perspectives in Dermatology. Biomedicines 2021, 9, 237. [Google Scholar] [CrossRef]
  30. Tao, R.E.; Prajapati, S.; Pixley, J.N.; Grada, A.; Feldman, S.R. Oral Tetracycline-Class Drugs in Dermatology: Impact of Food Intake on Absorption and Efficacy. Antibiotics 2023, 12, 1152. [Google Scholar] [CrossRef] [PubMed]
  31. Kim, Y.-S.; Kim, H.S. Tetracyclines Revisited: Tetracyclines in the Field of Dermatology. Dermatology 2024, 240, 1–15. [Google Scholar] [CrossRef]
  32. Bagatin, E.; Costa, C.S.; da Rocha, M.A.D.; Picosse, F.R.; Kamamoto, C.S.L.; Pirmez, R.; Ianhez, M.; Miot, H.A. Consensus on the Use of Oral Isotretinoin in Dermatology—Brazilian Society of Dermatology. An. Bras. Dermatol. 2020, 95, 19–38. [Google Scholar] [CrossRef]
  33. Kelhälä, H.; Fyhrquist, N.; Palatsi, R.; Lehtimäki, S.; Väyrynen, J.P.; Kubin, M.E.; Kallioinen, M.; Alenius, H.; Tasanen, K.; Lauerma, A. Isotretinoin Treatment Reduces Acne Lesions but Not Directly Lesional Acne Inflammation. Exp. Dermatol. 2016, 25, 477–478. [Google Scholar] [CrossRef]
  34. Bae, I.H.; Kwak, J.H.; Na, C.H.; Kim, M.S.; Shin, B.S.; Choi, H. A Comprehensive Review of the Acne Grading Scale in 2023. Ann. Dermatol. 2024, 36, 65–73. [Google Scholar] [CrossRef]
  35. Reynolds, R.V.; Yeung, H.; Cheng, C.E.; Cook-Bolden, F.; Desai, S.R.; Druby, K.M.; Freeman, E.E.; Keri, J.E.; Stein Gold, L.F.; Tan, J.K.L.; et al. Guidelines of Care for the Management of Acne Vulgaris. J. Am. Acad. Dermatol. 2024, 90, 1006.e1–1006.e30. [Google Scholar] [CrossRef]
  36. Auffret, N.; Nguyen, J.M.; Leccia, M.-T.; Claudel, J.P.; Dréno, B. TRASS: A Global Approach to Assess the Severity of Truncal Acne. J. Eur. Acad. Dermatol. Venereol. 2022, 36, 897–904. [Google Scholar] [CrossRef]
  37. Puig, L.; Guerra-Tapia, A.; Conejo-Mir, J.; Toribio, J.; Berasategui, C.; Zsolt, I. Validation of the Spanish Acne Severity Scale (Escala de Gravedad Del Acné Española—EGAE). Eur. J. Dermatol. 2013, 23, 233–240. [Google Scholar] [CrossRef]
  38. O’brien, S.; Lewis, J.; Cunliffe, W. The Leeds Revised Acne Grading System. J. Dermatol. Treat. 1998, 9, 215–220. [Google Scholar] [CrossRef]
  39. Dréno, B.; Poli, F.; Pawin, H.; Beylot, C.; Faure, M.; Chivot, M.; Auffret, N.; Moyse, D.; Ballanger, F.; Revuz, J. Development and Evaluation of a Global Acne Severity Scale (GEA Scale) Suitable for France and Europe. J. Eur. Acad. Dermatol. Venereol. 2011, 25, 43–48. [Google Scholar] [CrossRef] [PubMed]
  40. Bernardis, E.; Shou, H.; Barbieri, J.S.; McMahon, P.J.; Perman, M.J.; Rola, L.A.; Streicher, J.L.; Treat, J.R.; Castelo-Soccio, L.; Yan, A.C. Development and Initial Validation of a Multidimensional Acne Global Grading System Integrating Primary Lesions and Secondary Changes. JAMA Dermatol. 2020, 156, 296–302. [Google Scholar] [CrossRef] [PubMed]
  41. Tan, J.K.L.; Tang, J.; Fung, K.; Gupta, A.K.; Thomas, D.R.; Sapra, S.; Lynde, C.; Poulin, Y.; Gulliver, W.; Sebaldt, R.J. Development and Validation of a Comprehensive Acne Severity Scale. J. Cutan. Med. Surg. 2007, 11, 211–216. [Google Scholar] [CrossRef]
  42. O’Neill, A.M.; Cavagnero, K.J.; Seidman, J.S.; Zaramela, L.; Chen, Y.; Li, F.; Nakatsuji, T.; Cheng, J.Y.; Tong, Y.L.; Do, T.H.; et al. Genetic and Functional Analyses of Cutibacterium Acnes Isolates Reveal the Association of a Linear Plasmid with Skin Inflammation. J. Investig. Dermatol. 2024, 144, 116–124.e4. [Google Scholar] [CrossRef]
  43. Cobian, N.; Garlet, A.; Hidalgo-Cantabrana, C.; Barrangou, R. Comparative Genomic Analyses and CRISPR-Cas Characterization of Cutibacterium Acnes Provide Insights Into Genetic Diversity and Typing Applications. Front. Microbiol. 2021, 12, 758749. [Google Scholar] [CrossRef]
  44. Mina-Vargas, A.; Colodro-Conde, L.; Grasby, K.; Zhu, G.; Gordon, S.; Medland, S.E.; Martin, N.G. Heritability and GWAS Analyses of Acne in Australian Adolescent Twins. Twin Res. Hum. Genet. 2017, 20, 541–549. [Google Scholar] [CrossRef] [PubMed]
  45. Melnik, B.C. Acne Transcriptomics: Fundamentals of Acne Pathogenesis and Isotretinoin Treatment. Cells 2023, 12, 2600. [Google Scholar] [CrossRef] [PubMed]
  46. Paichitrojjana, A.; Paichitrojjana, A. Oral Isotretinoin and Its Uses in Dermatology: A Review. Drug Des. Devel. Ther. 2023, 17, 2573–2591. [Google Scholar] [CrossRef]
  47. Kabbani, D.; Akika, R.; Wahid, A.; Daly, A.K.; Cascorbi, I.; Zgheib, N.K. Pharmacogenomics in Practice: A Review and Implementation Guide. Front. Pharmacol. 2023, 14, 1189976. [Google Scholar] [CrossRef]
  48. Pirmohamed, M. Pharmacogenomics: Current Status and Future Perspectives. Nat. Rev. Genet. 2023, 24, 350–362. [Google Scholar] [CrossRef] [PubMed]
  49. Lin, Q.; Cai, B.; Ke, R.; Chen, L.; Ni, X.; Liu, H.; Lin, X.; Wang, B.; Shan, X. Integrative Bioinformatics and Experimental Validation of Hub Genetic Markers in Acne Vulgaris: Toward Personalized Diagnostic and Therapeutic Strategies. J. Cosmet. Dermatol. 2024, 23, 1777–1799. [Google Scholar] [CrossRef]
  50. Tan, I.J.; Podwojniak, A.; Parikh, A.; Cohen, B.A. Precision Dermatology: A Review of Molecular Biomarkers and Personalized Therapies. Curr. Issues Mol. Biol. 2024, 46, 2975–2990. [Google Scholar] [CrossRef]
  51. González, J.R.; Armengol, L.; Solé, X.; Guinó, E.; Mercader, J.M.; Estivill, X.; Moreno, V. SNPassoc: An R Package to Perform Whole Genome Association Studies. Bioinformatics 2007, 23, 654–655. [Google Scholar] [CrossRef]
  52. Loh, P.-R.; Tucker, G.; Bulik-Sullivan, B.K.; Vilhjálmsson, B.J.; Finucane, H.K.; Salem, R.M.; Chasman, D.I.; Ridker, P.M.; Neale, B.M.; Berger, B.; et al. Efficient Bayesian Mixed-Model Analysis Increases Association Power in Large Cohorts. Nat. Genet. 2015, 47, 284–290. [Google Scholar] [CrossRef]
  53. Shannon, P.; Markiel, A.; Ozier, O.; Baliga, N.S.; Wang, J.T.; Ramage, D.; Amin, N.; Schwikowski, B.; Ideker, T. Cytoscape: A Software Environment for Integrated Models of Biomolecular Interaction Networks. Genome Res. 2003, 13, 2498–2504. [Google Scholar] [CrossRef]
  54. Bindea, G.; Mlecnik, B.; Hackl, H.; Charoentong, P.; Tosolini, M.; Kirilovsky, A.; Fridman, W.-H.; Pagès, F.; Trajanoski, Z.; Galon, J. ClueGO: A Cytoscape Plug-in to Decipher Functionally Grouped Gene Ontology and Pathway Annotation Networks. Bioinformatics 2009, 25, 1091–1093. [Google Scholar] [CrossRef]
  55. Mann, O.N.; Kong, C.-S.; Lucas, E.S.; Brosens, J.J.; Hanyaloglu, A.C.; Brighton, P.J. Expression and Function of the Luteinizing Hormone Choriogonadotropin Receptor in Human Endometrial Stromal Cells. Sci. Rep. 2022, 12, 8624. [Google Scholar] [CrossRef]
  56. Sharma, S.; Shen, T.; Chitranshi, N.; Gupta, V.; Basavarajappa, D.; Sarkar, S.; Mirzaei, M.; You, Y.; Krezel, W.; Graham, S.L.; et al. Retinoid X Receptor: Cellular and Biochemical Roles of Nuclear Receptor with a Focus on Neuropathological Involvement. Mol. Neurobiol. 2022, 59, 2027–2050. [Google Scholar] [CrossRef] [PubMed]
  57. de Almeida, N.R.; Conda-Sheridan, M. A Review of the Molecular Design and Biological Activities of RXR Agonists. Med. Res. Rev. 2019, 39, 1372–1397. [Google Scholar] [CrossRef] [PubMed]
  58. Tsay, A.; Wang, J.-C. The Role of PIK3R1 in Metabolic Function and Insulin Sensitivity. Int. J. Mol. Sci. 2023, 24, 12665. [Google Scholar] [CrossRef] [PubMed]
  59. De Bortoli, M.; Queisser, A.; Pham, V.C.; Dompmartin, A.; Helaers, R.; Boutry, S.; Claus, C.; De Roo, A.-K.; Hammer, F.; Brouillard, P.; et al. Somatic Loss-of-Function PIK3R1 and Activating Non-Hotspot PIK3CA Mutations Associated with Capillary Malformation with Dilated Veins (CMDV). J. Investig. Dermatol. 2024, 144, 2066–2077.e6. [Google Scholar] [CrossRef]
  60. Wang, K.-Y.; Yamada, S.; Izumi, H.; Tsukamoto, M.; Nakashima, T.; Tasaki, T.; Guo, X.; Uramoto, H.; Sasaguri, Y.; Kohno, K. Critical in Vivo Roles of WNT10A in Wound Healing by Regulating Collagen Expression/Synthesis in WNT10A-Deficient Mice. PLoS ONE 2018, 13, e0195156. [Google Scholar] [CrossRef]
  61. Benard, E.L.; Hammerschmidt, M. The Fundamentals of WNT10A. Differentiation 2025, 142, 100838. [Google Scholar] [CrossRef]
  62. Jin, Z.; Song, Y.; He, L. A Review of Skin Immune Processes in Acne. Front. Immunol. 2023, 14, 1324930. [Google Scholar] [CrossRef]
  63. Cruz, S.; Vecerek, N.; Elbuluk, N. Targeting Inflammation in Acne: Current Treatments and Future Prospects. Am. J. Clin. Dermatol. 2023, 24, 681–694. [Google Scholar] [CrossRef] [PubMed]
  64. Tuchayi, S.M.; Makrantonaki, E.; Ganceviciene, R.; Dessinioti, C.; Feldman, S.R.; Zouboulis, C.C. Acne Vulgaris. Nat. Rev. Dis. Primers 2015, 1, 15029. [Google Scholar] [CrossRef] [PubMed]
  65. Grech, I.; Giatrakou, S.; Damoraki, G.; Pistiki, A.; Kaldrimidis, P.; Giamarellos-Bourboulis, E.J.; Stavrianeas, N. Single Nucleotide Polymorphisms of Toll-like Receptor-4 Protect against Acne Conglobata. J. Eur. Acad. Dermatol. Venereol. 2012, 26, 1538–1543. [Google Scholar] [CrossRef]
  66. Romics, L.; Dolganiuc, A.; Kodys, K.; Drechsler, Y.; Oak, S.; Velayudham, A.; Mandrekar, P.; Szabo, G. Selective Priming to Toll-like Receptor 4 (TLR4), Not TLR2, Ligands by P. Acnes Involves up-Regulation of MD-2 in Mice. Hepatology 2004, 40, 555–564. [Google Scholar] [CrossRef]
  67. Jugeau, S.; Tenaud, I.; Knol, A.C.; Jarrousse, V.; Quereux, G.; Khammari, A.; Dreno, B. Induction of Toll-like Receptors by Propionibacterium Acnes. Br. J. Dermatol. 2005, 153, 1105–1113. [Google Scholar] [CrossRef] [PubMed]
  68. Zhang, B.; Choi, Y.M.; Lee, J.; An, I.S.; Li, L.; He, C.; Dong, Y.; Bae, S.; Meng, H. Toll-like Receptor 2 Plays a Critical Role in Pathogenesis of Acne Vulgaris. Biomed. Dermatol. 2019, 3, 4. [Google Scholar] [CrossRef]
  69. Stefanovic, N.; Irvine, A.D. Filaggrin and Beyond. Ann. Allergy Asthma Immunol. 2024, 132, 187–195. [Google Scholar] [CrossRef]
  70. Sandilands, A.; Sutherland, C.; Irvine, A.D.; McLean, W.H.I. Filaggrin in the Frontline: Role in Skin Barrier Function and Disease. J. Cell Sci. 2009, 122, 1285–1294. [Google Scholar] [CrossRef]
  71. Common, J.E.A.; Brown, S.J.; Haines, R.L.; Goh, C.S.M.; Chen, H.; Balakrishnan, A.; Munro, C.S.; Tan, A.W.H.; Tan, H.H.; Tang, M.B.Y.; et al. Filaggrin Null Mutations Are Not a Protective Factor for Acne Vulgaris. J. Investig. Dermatol. 2011, 131, 1378–1380. [Google Scholar] [CrossRef]
  72. Hughes, A.J.; Barbosa, E.; Cernova, J.; Thomas, B.R.; O’Shaughnessy, R.F.L.; O’Toole, E.A. Loss-of-Function FLG Mutations Are Associated with Reduced History of Acne Vulgaris in a Cohort of Patients with Atopic Eczema of Bangladeshi Ancestry in East London. Clin. Exp. Dermatol. 2024, 49, 1547–1553. [Google Scholar] [CrossRef]
  73. Chen, X.; Min, S.; Chen, C.; Lin, X.; Wang, D.; Jiang, G. Influence of RETN, IL-1, and IL-6 Gene Polymorphisms on the Risk of Acne Vulgaris in the Chinese Population. J. Cosmet. Dermatol. 2022, 21, 4965–4973. [Google Scholar] [CrossRef]
  74. Lee, A.-Y. Molecular Mechanism of Epidermal Barrier Dysfunction as Primary Abnormalities. Int. J. Mol. Sci. 2020, 21, 1194. [Google Scholar] [CrossRef]
  75. Chen, Z.; Dragan, M.; Sun, P.; Haensel, D.; Vu, R.; Cui, L.; Zhu, P.; Yang, N.; Shi, Y.; Dai, X. The AhR-Ovol1-Id1 Regulatory Axis in Keratinocytes Promotes Epidermal and Immune Homeostasis in Atopic Dermatitis-like Skin Inflammation. Cell. Mol. Immunol. 2025, 22, 300–315. [Google Scholar] [CrossRef] [PubMed]
  76. Dragan, M.; Chen, Z.; Li, Y.; Le, J.; Sun, P.; Haensel, D.; Sureshchandra, S.; Pham, A.; Lu, E.; Pham, K.T.; et al. Ovol1/2 Loss-induced Epidermal Defects Elicit Skin Immune Activation and Alter Global Metabolism. EMBO Rep. 2023, 24, e56214. [Google Scholar] [CrossRef] [PubMed]
  77. Sun, P.; Vu, R.; Dragan, M.; Haensel, D.; Gutierrez, G.; Nguyen, Q.; Greenberg, E.; Chen, Z.; Wu, J.; Atwood, S.; et al. OVOL1 Regulates Psoriasis-Like Skin Inflammation and Epidermal Hyperplasia. J. Investig. Dermatol. 2021, 141, 1542–1552. [Google Scholar] [CrossRef] [PubMed]
Figure 1. Manhattan-style plot of −log10(p) for the 60-SNP panel evaluated under five inheritance models (additive [primary], codominant, dominant, recessive, overdominant). The x-axis lists SNPs in panel order, and the y-axis shows −log10(p) from single-variant tests. The blue dashed line marks the nominal threshold (p = 0.05), and the red dashed line marks the Bonferroni threshold for 60 tests (p = 0.05/60). Variants crossing the Bonferroni line include, in the additive model: rs28988604 (CYP3A4), rs10515088 (PIK3R1), rs10918169 (RXR), rs13429458 (THADA), rs1159268 (TGF-β2), rs38055 (FST), and rs28371686 (CYP2C9); under alternative models: rs74333950 (WNT10A, recessive), rs13405728 (LHCGR, recessive), rs1057910 (CYP2C9, recessive), and rs2650000 (HNF1A-AS1, recessive).
Figure 1. Manhattan-style plot of −log10(p) for the 60-SNP panel evaluated under five inheritance models (additive [primary], codominant, dominant, recessive, overdominant). The x-axis lists SNPs in panel order, and the y-axis shows −log10(p) from single-variant tests. The blue dashed line marks the nominal threshold (p = 0.05), and the red dashed line marks the Bonferroni threshold for 60 tests (p = 0.05/60). Variants crossing the Bonferroni line include, in the additive model: rs28988604 (CYP3A4), rs10515088 (PIK3R1), rs10918169 (RXR), rs13429458 (THADA), rs1159268 (TGF-β2), rs38055 (FST), and rs28371686 (CYP2C9); under alternative models: rs74333950 (WNT10A, recessive), rs13405728 (LHCGR, recessive), rs1057910 (CYP2C9, recessive), and rs2650000 (HNF1A-AS1, recessive).
Cosmetics 12 00217 g001
Figure 2. Gene–gene interaction network from 190 significant Single-Nucleotide Polymorphism (SNP) pairs (FDR ≤ 0.10). Node size indicates degree (number of significant interactions). Edge thickness reflects statistical weight [−log10(p)] of the top supporting SNP pair per gene–gene edge. TLR4 is the principal hub connecting immune, lipid, barrier–keratinisation, and endocrine modules.
Figure 2. Gene–gene interaction network from 190 significant Single-Nucleotide Polymorphism (SNP) pairs (FDR ≤ 0.10). Node size indicates degree (number of significant interactions). Edge thickness reflects statistical weight [−log10(p)] of the top supporting SNP pair per gene–gene edge. TLR4 is the principal hub connecting immune, lipid, barrier–keratinisation, and endocrine modules.
Cosmetics 12 00217 g002
Figure 3. TLR4-centred interaction map grouped by functional pathways. The central red node is TLR4. Blue nodes are pathway categories derived from the interaction network and ClueGO grouping (innate/adaptive immunity; lipid metabolism; epidermal barrier–keratinisation; endocrine/gonadal axis). Green nodes list representative interacting genes from the selected SNP–SNP pairs within each pathway. Grey edges connect TLR4 → pathway and pathway → genes (visual summary; edge thickness not scaled).
Figure 3. TLR4-centred interaction map grouped by functional pathways. The central red node is TLR4. Blue nodes are pathway categories derived from the interaction network and ClueGO grouping (innate/adaptive immunity; lipid metabolism; epidermal barrier–keratinisation; endocrine/gonadal axis). Green nodes list representative interacting genes from the selected SNP–SNP pairs within each pathway. Grey edges connect TLR4 → pathway and pathway → genes (visual summary; edge thickness not scaled).
Cosmetics 12 00217 g003
Figure 4. Bar graph showing the top 12 Gene Ontology (GO) biological processes identified by ClueGO after excluding broad reproductive categories. Enrichment significance is displayed as −log10(p) for readability. The leading processes—lipid localisation and transport, keratinisation and epidermal differentiation, inflammatory regulation including cytokine signalling, nitric oxide biosynthesis, and endocrine development—highlight the convergence of barrier, metabolic, and immune mechanisms in severe acne. Analysis parameters: GO Biological Process in ClueGO/CluePedia with the whitelist-mapped genes as background; multiple testing controlled by Benjamini–Hochberg false discovery rate (FDR). Inference was based on q values, and enrichments are interpreted as suggestive under exploratory thresholds.
Figure 4. Bar graph showing the top 12 Gene Ontology (GO) biological processes identified by ClueGO after excluding broad reproductive categories. Enrichment significance is displayed as −log10(p) for readability. The leading processes—lipid localisation and transport, keratinisation and epidermal differentiation, inflammatory regulation including cytokine signalling, nitric oxide biosynthesis, and endocrine development—highlight the convergence of barrier, metabolic, and immune mechanisms in severe acne. Analysis parameters: GO Biological Process in ClueGO/CluePedia with the whitelist-mapped genes as background; multiple testing controlled by Benjamini–Hochberg false discovery rate (FDR). Inference was based on q values, and enrichments are interpreted as suggestive under exploratory thresholds.
Cosmetics 12 00217 g004
Figure 5. Violin plots of standardised Polygenic Risk Score (PRS) across acne severity grades (1, 2–3, 4). Distributions show higher median scores and greater spread with increasing severity.
Figure 5. Violin plots of standardised Polygenic Risk Score (PRS) across acne severity grades (1, 2–3, 4). Distributions show higher median scores and greater spread with increasing severity.
Cosmetics 12 00217 g005
Figure 6. Receiver operating characteristic (ROC) curves for discrimination between Grade 4 and Grade 1 (blue) and between Grade 4 and Grade 2 (orange), using the predicted probability of Grade 4 as the score.
Figure 6. Receiver operating characteristic (ROC) curves for discrimination between Grade 4 and Grade 1 (blue) and between Grade 4 and Grade 2 (orange), using the predicted probability of Grade 4 as the score.
Cosmetics 12 00217 g006
Figure 7. Receiver operating characteristic (ROC) curve for discrimination between Grade 4 and the combined Grades 1 and 2, using the predicted probability of Grade 4 as the score.
Figure 7. Receiver operating characteristic (ROC) curve for discrimination between Grade 4 and the combined Grades 1 and 2, using the predicted probability of Grade 4 as the score.
Cosmetics 12 00217 g007
Figure 8. Density plots of standardised Polygenic Risk Score (PRS) stratified by acne severity grade, illustrating the rightward shift in distribution for severe cases.
Figure 8. Density plots of standardised Polygenic Risk Score (PRS) stratified by acne severity grade, illustrating the rightward shift in distribution for severe cases.
Cosmetics 12 00217 g008
Table 1. List of the 60 single-nucleotide polymorphisms (SNPs) included in the dataset with their respective gene annotations, as well as the corresponding major and minor alleles with their frequencies. The variants were selected based on evidence from the literature and prior internal studies that linked them to biological pathways relevant to acne pathophysiology. These pathways include inflammation, sebaceous gland activity, follicular keratinisation, androgen signalling and lipid metabolism. Selection criteria ensured that the panel covered both common variants with established roles and pharmacogenetic SNPs previously studied in relation to acne-related drug response, although pharmacogenetic associations were not analysed in the present work.
Table 1. List of the 60 single-nucleotide polymorphisms (SNPs) included in the dataset with their respective gene annotations, as well as the corresponding major and minor alleles with their frequencies. The variants were selected based on evidence from the literature and prior internal studies that linked them to biological pathways relevant to acne pathophysiology. These pathways include inflammation, sebaceous gland activity, follicular keratinisation, androgen signalling and lipid metabolism. Selection criteria ensured that the panel covered both common variants with established roles and pharmacogenetic SNPs previously studied in relation to acne-related drug response, although pharmacogenetic associations were not analysed in the present work.
rsIDGene or RegionMajor AlleleMinor AlleleSNP Class
rs717620ABCC2TCPharmacogenetic/PK—efflux transporter (xenobiotic/retinoid handling).
rs6544713ABCG8CTLipid/Metabolic—sterol transport; links to sebum lipid composition.
rs4420638APOEGALipid/Metabolic + Immune—apolipoprotein; lipid–inflammation crosstalk.
rs9667947ARAP1TCEndocrine/Metabolic—insulin signalling; metabolic influence on sebaceous biology.
rs2212434C11orf30/LRRC32TCImmune Regulation—GARP/TGF-β axis; Treg-related.
rs743572CYP17A1AGEndocrine/Androgen—CYP17A1 steroidogenesis (androgen pathway).
rs700518CYP19ACTEndocrine/Androgen—CYP19A1 aromatase (androgen–oestrogen balance).
rs1799853CYP2C9*2CTPharmacogenetic/PK—CYP2C9*2 drug metabolism; retinoid biotransformation.
rs1057910CYP2C9*3CAPharmacogenetic/PK—CYP2C9*3 drug metabolism; retinoid biotransformation.
rs28371686CYP2C9*5ACPharmacogenetic/PK—CYP2C9*5 metabolic variability.
rs7900194CYP2C9*8CTPharmacogenetic/PK—CYP2C9*8 metabolic variability.
rs28988604CYP3A4*11AGPharmacogenetic/PK—CYP3A4*11 retinoid metabolism.
rs2737418CYP3A4*2TCPharmacogenetic/PK—CYP3A4*2 retinoid metabolism.
rs67666821CYP3A4*20CTPharmacogenetic/PK—CYP3A4*20 retinoid metabolism.
rs35599367CYP3A4*22GAPharmacogenetic/PK—CYP3A4*22 reduced expression/activity.
rs776746CYP3A5CTPharmacogenetic/PK—CYP3A5*3 splicing variant; CYP3A activity.
rs1799883FABP2CTLipid/Metabolic—intestinal FA handling; systemic lipid flux.
rs7927894FLGCTBarrier–keratinisation—FLG; epidermal barrier integrity.
rs12123821FLG-AS1CTBarrier–keratinisation—FLG-AS1; barrier–immune regulation.
rs1511412FOXL2GAEndocrine/Regulatory—FOXL2 transcription factor; hormonal axis.
rs2268361FSHRCTEndocrine/Androgen—FSHR signalling; androgen milieu.
rs2349415FSHRTCEndocrine/Androgen—FSHR signalling (second marker at locus).
rs38055FSTAGImmune/Remodelling—FST (activin/TGF-β regulator).
rs8050136FTOCALipid/Metabolic—adiposity axis (FTO); sebum link.
rs27647GHRLCTEndocrine/Metabolic—ghrelin signalling; sebaceous activity.
rs2844573HLA-B*13:01ACImmune/HLA—HLA-B*13:01 antigen presentation; immunogenetics/ADR context.
rs2442736HLA-B*51:01CGImmune/HLA—HLA-B*51:01 antigen presentation; inflammatory risk.
rs763035HLA-DRAGAImmune/HLA—HLA-DRA antigen presentation.
rs701829HLA-DRB1CTImmune/HLA—HLA-DRB1 antigen presentation.
rs2650000HNF1A-AS1ACTranscription/Metabolic—HNF1A-AS1; hepatic lipid regulation.
rs1800896IL-10TCImmune/Cytokine—IL10 anti-inflammatory signalling.
rs1295686IL-13CTImmune/Cytokine—IL13 Th2 milieu.
rs16944IL-1BGAImmune/Cytokine—IL1B pro-inflammatory signalling.
rs12203592IRF4CTImmune/Transcription—IRF4 lymphocyte differentiation.
rs13405728LHCGRGAEndocrine/Androgen—LHCGR; androgenic influence on sebocytes.
rs17030203MTA3TGTranscription/Epithelial—MTA3 (ER co-repressor); differentiation.
rs4133274MYCAGTranscription/Proliferation—MYC; keratinocyte/sebocyte signalling.
rs1426654MYEF2CTTranscription/Regulatory—MYEF2; immune/epithelial context.
rs873549Non-genic regionTCExploratory/Intergenic—tag/regulatory signal from prior evidence.
rs4149056OATP1B1CTPharmacogenetic/PK—SLCO1B1 (OATP1B1) hepatic uptake; systemic handling.
rs7103693ODZ4CTTranscription/Regulatory—ODZ4/TENM4; developmental/regulatory candidate.
rs478304OVOL1GTBarrier–keratinisation—OVOL1; epidermal differentiation TF.
rs10515088PIK3R1GAPI3K Signalling—PIK3R1; inflammation/sebocyte biology.
rs738409PNPLA3GCLipid/Metabolic—PNPLA3 lipid droplet/remodelling; sebum composition.
rs1862513RETNCGLipid/Metabolic + Inflammation—RETN (resistin) adipokine.
rs3745367RETNGALipid/Metabolic + Inflammation—RETN (second marker at locus).
rs10918169RXRCGRetinoid Signalling—RXR nuclear receptor (pharmacodynamic axis).
rs1128977RXRAGRetinoid Signalling—RXR nuclear receptor.
rs2651860RXRCARetinoid Signalling—RXR nuclear receptor.
rs283696RXRTCRetinoid Signalling—RXR nuclear receptor.
rs12964116SERPINB7GABarrier–keratinisation—SERPINB7; epidermal protease inhibition.
rs404818SOAT1TCLipid/Metabolic—SOAT1/ACAT1 cholesterol esterification in sebocytes.
rs1159268TGF-β2AGImmune/Remodelling—TGF-β2; matrix turnover/scarring.
rs12478601THADATCEndocrine/Metabolic—THADA (insulin/thyroid-linked biology).
rs13429458THADACAEndocrine/Metabolic—THADA (second marker at locus).
rs4986790TLR4GAInnate Immunity—TLR4; C. acnes recognition (hub).
rs4986791TLR4TCInnate Immunity—TLR4 (co-segregating functional variant).
rs58542926TM6SF2CTLipid/Metabolic—TM6SF2 hepatic TG secretion; sebum lipid profile.
rs1800629TNF-αGAImmune/Cytokine—TNF-α pro-inflammatory signalling.
rs74333950WNT10AGTBarrier/Development—WNT10A follicular/epidermal morphogenesis.
Table 2. Single-Nucleotide Polymorphisms (SNPs) retained after relaxed quality control thresholds and used in Polygenic Risk Score (PRS) and SNP–SNP interaction analyses. MAF—Minor allele frequency; HWE—Hardy–Weinberg equilibrium.
Table 2. Single-Nucleotide Polymorphisms (SNPs) retained after relaxed quality control thresholds and used in Polygenic Risk Score (PRS) and SNP–SNP interaction analyses. MAF—Minor allele frequency; HWE—Hardy–Weinberg equilibrium.
SNPCall RateMAFHWE p-Value
rs105150881.0000.03461.06 × 10−36
rs11592681.0000.09461.30 × 10−10
rs121238211.0000.45777.01 × 10−50
rs122035921.0000.11851.82 × 10−25
rs124786011.0000.15235.21 × 10−7
rs12956861.0000.11003.70 × 10−4
rs129641161.0000.42316.61 × 10−8
rs134057280.98770.12079.95 × 10−9
rs134294581.0000.06158.82 × 10−14
rs14266541.0000.37310.737
rs15114121.0000.46000.0698
rs169441.0000.26380.00631
rs170302031.0000.31542.47 × 10−22
rs17998830.98770.15110.6455
rs18006291.0000.29920.6399
rs18008961.0000.19770.3864
rs18625131.0000.12770.05224
rs22124341.0000.15770.8824
rs22683611.0000.41000.00267
rs23494150.98770.39250.9341
rs26500001.0000.16085.20 × 10−12
rs276471.0000.20154.88 × 10−46
rs37453671.0000.23853.39 × 10−4
rs380550.98770.34420.000627
rs4048181.0000.10926.17 × 10−11
rs41332740.98770.21422.63 × 10−26
rs41490561.0000.093851.01 × 10−26
rs44206381.0000.24310.00535
rs4783041.0000.27310.02323
rs49867901.0000.43466.89 × 10−7
rs49867911.0000.36540.09056
rs585429261.0000.18920.01000
rs65447131.0000.051540.4050
rs7005181.0000.35310.4929
rs7018291.0000.41310.7464
rs71036931.0000.085382.53 × 10−49
rs7176201.0000.48080.03379
rs7384090.97230.16936.53 × 10−25
rs743339501.0000.26380.2266
rs7435721.0000.14006.47 × 10−64
rs79278941.0000.31690.9278
rs80501361.0000.36157.22 × 10−8
rs8735491.0000.34380.00310
rs96679471.0000.22776.78 × 10−20
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Russo, V.; Vila-Vecilla, L.; Guerrero, A.S.; Madrigal, L.G.; Chiovatto, C.B.; de Souza, G.T. Single-Locus, Interaction, and Functional Pathway Analyses of Acne Severity in a 60-SNP Panel. Cosmetics 2025, 12, 217. https://doi.org/10.3390/cosmetics12050217

AMA Style

Russo V, Vila-Vecilla L, Guerrero AS, Madrigal LG, Chiovatto CB, de Souza GT. Single-Locus, Interaction, and Functional Pathway Analyses of Acne Severity in a 60-SNP Panel. Cosmetics. 2025; 12(5):217. https://doi.org/10.3390/cosmetics12050217

Chicago/Turabian Style

Russo, Valentina, Laura Vila-Vecilla, Albert Sanchez Guerrero, Laura Gascón Madrigal, Caroline Brandão Chiovatto, and Gustavo Torres de Souza. 2025. "Single-Locus, Interaction, and Functional Pathway Analyses of Acne Severity in a 60-SNP Panel" Cosmetics 12, no. 5: 217. https://doi.org/10.3390/cosmetics12050217

APA Style

Russo, V., Vila-Vecilla, L., Guerrero, A. S., Madrigal, L. G., Chiovatto, C. B., & de Souza, G. T. (2025). Single-Locus, Interaction, and Functional Pathway Analyses of Acne Severity in a 60-SNP Panel. Cosmetics, 12(5), 217. https://doi.org/10.3390/cosmetics12050217

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop