Glossary

AhR (Aryl hydrocarbon receptor): Environmental sensor in skin cells; modulates barrier function, immunity, and oxidative-stress response. Microbiome-derived indole metabolites are key ligands. (Ch5, Ch6)

AlphaFold (2 / 3): DeepMind's protein structure prediction AI. AlphaFold2 (2021) achieved mean GDT-TS 87.0 at CASP14, essentially solving single-chain structure prediction. AlphaFold3 (2024) extends to protein-ligand complexes. (Ch5)

Antimicrobial peptide (AMP): Short peptides that disrupt microbial membranes. Both skin-derived and microbe-derived AMPs help maintain community homeostasis. (Ch2, Ch4)

Bayesian optimization: Active-learning technique that minimizes expensive experimental evaluations while exploring multi-objective combinatorial spaces — the workhorse of AI formulation. (Ch8)

Boltz-1: MIT's open-source AlphaFold3 equivalent released November 2024. Permissive licensing lowers cosmetic R&D entry barriers. (Ch5)

Chai-1: Another AlphaFold3 open alternative. Joint multi-modal molecular interaction prediction. (Ch5)

Clinical simulation: In silico prediction of clinical outcomes. PBPK/QSP techniques imported from pharma now apply to microbiome actives. (Ch9)

Cosmetogenomics: Evaluation and design of cosmetic efficacy using genomic / metagenomic data. (Ch2, Ch4)

Culturomics: Systematic isolation and identification of culturable microbes. Complements NGS and provides ground truth for synbio and clinical validation. (Ch3, Ch7)

Cutibacterium acnes: Gram-positive bacterium dominant in sebaceous skin sites (formerly *Propionibacterium*). Both commensal and acne-pathogenic. (Ch2)

DBTL loop (Design–Build–Test–Learn): Standard synthetic biology workflow. AI adds the most value at the Design and Learn stages. (Ch7)

Digital twin: Virtual model of a patient/consumer. Skin twins and microbiome twins are central to digital clinical trials and personalized formulation. (Ch6)

Dysbiosis: Microbiome imbalance. The definition remains fuzzy; an operationalizable reference state has not been established. (Ch2)

EPI-7: A novel *Epidermidibacterium keratini* strain (Aekyung/Encos R&D). Clinical efficacy in atopic skin reported in Kim 2023 *Int J Mol Sci*. (Ch1, Ch9, Ch10)

ESMFold / ESM3: Meta's protein language model–based structure prediction and generation models. Fast single-sequence inference. (Ch5)

Ex vivo skin model: Laboratory-maintained human skin tissue (EpiSkin, EpiDerm, full-thickness reconstructed human epidermis, etc.). (Ch9)

FACE-LINK: A 950-participant Korean skin microbiome cohort built by COSMAX × Dankook (Mun 2025 *Front Cell Infect Microbiol*). (Ch4, Ch10)

Fermentate / Ferment filtrate: Lactic acid bacterium / yeast fermentation broth. Core raw material of first-generation fermentation cosmetics. (Ch1)

Foundation model: A general-purpose model pretrained via self-supervision on broad data. No skin-microbiome-scale ESM3 equivalent exists yet. (Ch6, Ch12)

Generative AI: VAE / diffusion / transformer-based generative models, used for protein, small-molecule, and formulation generation. (Ch4, Ch5, Ch8)

GNN (Graph Neural Network): Neural networks that learn from graph-structured data such as microbial communities or molecular graphs. (Ch6)

HMP (Human Microbiome Project): NIH-led human-body microbiome mapping initiative, 2007–2013. (Ch2, Ch3)

Host-microbe interaction: Molecular signaling between host (skin) cells and microbes. Major pathways include TLR2/3, AhR, NF-κB. (Ch2, Ch5, Ch6)

iHSMGC (integrated Human Skin Microbial Gene Catalog): Li 2021 *Microbiome*. 10.93M genes, of which 4.88M (44.6%) were novel in an East Asian cohort — evidence of Western-biased reference data. (Ch2, Ch3, Ch10, Ch12)

In silico: Computer-simulation-based experimentation. (Ch5, Ch9)

Insilico Medicine: AI drug-discovery company. GENTRL → rentosertib (30-month target-to-Phase-1) → 2025 Phase 2a (FVC +98.4 mL vs −20.3 mL placebo) → Hong Kong IPO. (Ch11)

Lactobio: Danish microbiome IP / strain company (~10,000 strains) acquired by L'Oréal in 2023. (Ch4, Ch10)

Lactobacillus / Lactiplantibacillus: Lactic acid bacteria. Core resource of first-generation fermentation cosmetics. (Ch1, Ch7)

Malassezia: Skin-dominant lipid-dependent yeast. Implicated in seborrheic dermatitis and Malassezia folliculitis. (Ch2)

Metagenomics: Comprehensive sequencing and analysis of DNA from environmental samples (16S rRNA / shotgun / long-read). (Ch3)

MMP (Matrix Metalloproteinase): Enzyme family that degrades extracellular matrix (collagen, elastin). Central target of anti-aging cosmetics. (Ch5)

Multi-omics: Integration of multiple data layers — genome, transcriptome, metabolome, proteome. (Ch6)

NF-κB: Master transcription factor of inflammation and immunity. Target of anti-inflammatory cosmetics. (Ch5)

NGS (Next-Generation Sequencing): Successor to Sanger sequencing. Cost-per-genome dropped ~10^6× since 454 Life Sciences (2005). (Ch3)

NPAtlas / MIBiG / GNPS: Public databases of natural products, secondary metabolites, and biosynthetic gene clusters. Training corpora for AI screening. (Ch4)

PBPK / QSP (Physiologically-Based Pharmacokinetics / Quantitative Systems Pharmacology): Pharmacokinetic and systems-pharmacology modeling. Adapted to microbiome-active clinical simulation. (Ch9)

Postbiotic: Non-living microbial preparations or microbial metabolites. The mainstream cosmetic path given regulatory difficulty of live microbes. (Ch7)

Prebiotic: Substrates that promote the growth of beneficial microbes. (Ch7)

Retrosynthesis AI: AI that reasons backward from target molecules to starting materials (ASKCOS, AiZynthFinder, Chemformer). (Ch7)

RoseTTAFold (-AllAtom): Baker lab's protein structure prediction model — contemporaneous competitor to AlphaFold2. (Ch5)

Sebaceous / Moist / Dry: Grice & Segre's 2009 three-zone framework for skin sites. The reference frame for cosmetic efficacy claims. (Ch2)

Seres SER-109 (Vowst): First FDA-approved microbiome drug (2023). Recurrent *C. difficile* infection (rCDI) recurrence 12% vs placebo 40% in ECOSPOR III. (Ch7, Ch11)

Shotgun metagenomics: Whole-DNA sequencing (vs. 16S amplicon). Provides functional and strain-level resolution. (Ch3)

Skin microbiome: The community of bacteria, fungi, and viruses inhabiting the skin. (Ch2)

Spermidine: A polyamine. COSMAX-Dankook 2021 *Communications Biology* showed *Streptococcus*-secreted spermidine restores human dermal structure and barrier function. (Ch1, Ch5, Ch9)

Staphylococcus epidermidis: Gram-positive skin commensal. LTA → TLR2 signaling suppresses TLR3-mediated inflammation. (Ch2)

Stratum corneum: Outermost layer of dead skin cells; the primary barrier to cosmetic penetration. (Ch2, Ch8)

Synthetic biology: Bioengineering through genome design, DNA synthesis, and strain engineering. (Ch7)

Vedanta Biosciences: Microbiome therapeutics company built around defined microbial consortia. (Ch11)

Vowst: Seres' trade name for SER-109. See Seres SER-109. (Ch7, Ch11)