Case study · Bioinformatics · Dermo-cosmetics

A dermo-cosmetic active reshapes the wound-healing transcriptome of skin explants

RNA-seq analysis of 24 ex vivo human skin explants comparing an untreated condition to a dermo-cosmetic active, from raw reads to functional biological interpretation.

Model: ex vivo skin explant Technique: bulk RNA-seq n = 24 explants Genome: GRCh38.p14 / GENCODE v43

01Quality control

Before any biological comparison, every explant is screened on sequencing depth, alignment rate, duplication and base quality: split by condition to rule out a technical explanation for any downstream signal.

Percentage of mapped reads per condition, STAR 2-pass alignment
% reads mapped per condition: STAR 2-pass, GRCh38.p14 / GENCODE v43.
Distribution of QC metrics per condition: duplication, mapped reads, Q30
Duplication, % mapped, % Q30: both conditions within acceptable ranges.

02Analysis pipeline

A reproducible pipeline from alignment to functional interpretation, every tool and parameter version-controlled and documented for audit and reuse.

STAR 2-passDESeq2 (VST) apeglm shrinkageclusterProfiler 4.8.1 GO ORAKEGG pathway ORA

03Principal component analysis

PCA on DESeq2 VST-normalized counts confirms that the treatment condition, not batch or technical variation, is the dominant source of variance, with clean separation between groups before any statistical test is run.

PCA plot showing separation between treated and untreated skin explants
PC1 (41.2%) + PC2 (18.9%), separation driven by exposure to the dermo-cosmetic active.

04Differential expression

Treated vs. untreated comparison at |log2FC| > 1.5, padj < 0.05 (Benjamini-Hochberg) surfaces a coherent wound-healing signature: extracellular matrix and re-epithelialization genes induced, inflammatory and matrix-degrading genes reduced.

Volcano plot of differential expression, treated vs untreated skin explants
Volcano plot, treated vs. untreated, apeglm log2 fold change.
↑ COL1A1, COL3A1, FN1, TGFB1, POSTN, VEGFA, HAS2, KRT16/17 ↓ IL6, CXCL8, CCL2, MMP1, MMP3, MMP9

05Visualization, expression heatmap

Unsupervised clustering on the wound-healing gene panel separates the two conditions cleanly, with collagen, growth factor and re-epithelialization genes up in the treated group, and inflammatory/matrix-degrading genes up in the untreated group.

Heatmap of wound-healing genes across treated and untreated skin explants
Wound-healing gene panel, hierarchical clustering by sample and gene.

06Functional enrichment

Genes induced by the active map onto extracellular matrix organization, wound healing and keratinocyte/epithelial migration: with KEGG pathway analysis converging on ECM-receptor interaction, focal adhesion and the TGF-beta / PI3K-Akt signaling axes.

GO Biological Process over-representation analysis
GO Biological Process ORA, genes induced by the active.
KEGG pathway enrichment dotplot
KEGG pathway enrichment: ORA, BH-corrected.

07Methods & reproducibility

Alignment (STAR 2-pass, GRCh38.p14 / GENCODE v43), normalization and differential expression (DESeq2, apeglm shrinkage), and functional enrichment (clusterProfiler 4.8.1, GO + KEGG ORA with Benjamini-Hochberg correction) are documented with exact versions and thresholds, so results can be audited or reproduced by your own team.

Want the full report?

See the complete methodology, sample-level tables and gene lists.

Request the full case study