A Skin Microbiome–Inflammation Interaction Framework for Acne Severity Prediction

Thesis Abstract

The rising prevalence of inflammatory acne and interindividual variability in severity despite similar clinical presentations underscore the need to understand how skin microbial communities interact with host inflammatory responses to drive disease trajectories. This study addresses the problem that current predictive models of acne severity largely overlook dynamic microbiome–host immune interactions and their temporal evolution, limiting personalized management strategies. The aim is to develop a Skin Microbiome–Inflammation Interaction Framework capable of predicting acne severity by integrating microbial community structure, function, and host inflammatory markers. Specific objectives include (1) characterizing skin microbiome composition and inferred functional profiles across facial sebaceous zones in individuals with mild, moderate, and severe acne (n = 180; 60 per severity group); (2) quantifying local and systemic inflammatory signatures through targeted cytokine panels (IL-1?, IL-6, TNF-?, IL-17A) and antimicrobial peptide (?-defensin) levels, and correlating these with microbiome features; (3) developing and validating a predictive model that combines microbiome metrics (alpha/beta diversity, differential abundance of Cutibacterium acnes phylotypes, functional pathways inferred via PICRUSt2) with inflammatory biomarkers to forecast 12-week acne severity trajectories; and (4) evaluating the theoretical integration of the Medical Ecological Theory and the Skin–Mimmune Interaction Model to interpret how microbial–immune feedback loops modulate disease progression. The study employs a longitudinal, observational design with a diverse cohort recruited from dermatology clinics, ensuring representation across age, sex, and skin phototypes. Data collection comprises noninvasive stratified skin swabs from forehead, cheek, and chin for metagenomic sequencing (shotgun approach) and 16S rRNA gene profiling, combined with dermal surface impedance measurements and standardized sebum flow assessments. Serum inflammatory markers are assayed using multiplex ELISA, and dermatological severity is scored weekly using the Global Acne Grading System (GAGS) over 12 weeks. Data analysis adopts a multimodal approach (i) bioinformatics pipelines (MetaPhlAn3, HUMAnN3) to derive taxonomic and functional profiles, (ii) mixed-effects regression to model changes in severity as a function of microbial and inflammatory variables while accounting for repeated measures, (iii) machine learning ensemble methods (Elastic Net, Random Forest, and XGBoost) to construct and validate the predictive framework, and (iv) mediation analyses to test whether inflammatory markers mediate the relationship between microbiome features and severity outcomes. External validation will be performed with an independent cohort (n = 60). The expected findings include robust associations between specific Cutibacterium acnes phylotypes, functional pathways related to porphyrin metabolism and lipase activity, and heightened IL-1? and IL-6 levels with greater acne severity; the predictive model is anticipated to achieve an area under the receiver operating characteristic curve (AUC) exceeding 0.85 for predicting progression to moderate/severe acne within 12 weeks, using baseline microbiome and inflammatory profiles. The study contributes to knowledge by formalizing a frame for integrating microbial ecology with host immunology into predictive dermatology, advancing theories of skin microbial–immune co-dynamics and offering a transferable framework for other inflammatory skin conditions. Practical implications include informing personalized therapeutic strategies that target microbiome composition (e.g., phylotype-specific probiotics or bacteriophage approaches) and host inflammatory pathways, as well as guiding monitoring protocols in clinical practice. The main conclusion envisaged is that acne severity progression is substantially governed by interaction effects between microbiome function and inflammatory responses, rather than by single-factor determinants, and that a combined microbiome–inflammation model substantially improves predictive accuracy over models using either domain alone. Recommendations emphasize incorporating microbiome and inflammatory assessments into routine clinical research, exploring causal interventional studies to modulate identified microbial–immune axes, and extending the framework to longitudinal monitoring in diverse populations to ensure generalizability.