Principal Component Analysis

PCA, also known as principal component analysis or PCA, is a statistical-mathematical procedure that can be used to analyze and understand very large genetic data sets. Principal Component Analysis is a dimension-reducing statistical method for analyzing high-dimensional, genetic, transcriptomic or epigenetic data sets, with the help of which the main pattern of biological variation can be identified and visualized. PCA is used to:

• multiple genes
• multiple SNPs
• methylation sites
• transcriptome data

to a few meaningful dimensions.

Genetic data is extremely complex, e.g. 20,000 genes per sample, many genes correlate with each other and the differences are often subtle. Using appropriate algorithms, a common pattern can be analyzed. PCA can be used to recognize the main axes of variation in the data.

These are called:

• PC1 = largest genetic variance
• PC2 = second largest variance
• PC3 etc.

Each "component" is made up of a combination of many genes or SNPs.

One often finds:

• PC1 on the x-axis
• PC2 on the y-axis

Samples with similar genetics are close to each other, different groups are separated.

Applications in genetic engineering

population genetics

• Genetic origin
• Clusters (e.g. European / Asian / African)
• Control of population bias

RNA sequencing

Among other things, this enables the separation of:

• sick vs. healthy
• therapy responders vs. non-responders
• Quality control ("batch effects")

epigenetics

• DNA methylation profiles
• epigenetic ageing
• Tumor subtypes

Examples: Dermatology / Immunology: In inflammatory diseases (e.g. psoriasis vs. atopic dermatitis) distinct genetic profiles can be identified Th1/Th17-dominated patterns(IFN signature vs. IL4/IL13 signature).

Principal component analysis (PCA) shows different molecular expression patterns in atopic dermatitis and psoriasis, e.g.

• IL17A (psoriasis high/atopic dermatitis low)
• IFNG (psoriasis medium/atopic dermatitis low)
• IL4 (psoriasis low/atopic dermatitis high)
• IL13 (psoriasis low/atopic dermatitis high)

PCA thus automatically separates a gene sample into two clusters that can be assigned from the respective known data. PCA is therefore not a causal analysis but merely an exploratory analysis. PCA shows structures but not causes.

PCA algorithms and implementations can of course also be used with large scRNA-seq datasets (Tsuyuzaki K et al. 2020).

1. Ben Salem K et al.(2021) Principal Component Analysis (PCA). Tunis Med 99:383-389.
2. Moldovan LI et al. (2021) Characterization of circular RNA transcriptomes in psoriasis and atopic dermatitis reveals disease-specific expression profiles. Exp Dermatol 30:1187-1196.
3. Pardo LM et al. (2020) Principal component analysis of seven skin-ageing features identifies three main types of skin ageing. Br J Dermatol 182:1379-1387.
4. Traks T et al. (2024) High-throughput proteomic analysis of chronic inflammatory skin diseases: Psoriasis and atopic dermatitis. Exp Dermatol 33:e15079.
5. Tsuyuzaki K et al.(2020) Benchmarking principal component analysis for large-scale single-cell RNA-sequencing. Genome Biol 21:9.