Application of t-SNE to human genetic data

Area of Science:

• Genetics
• Bioinformatics
• Data Visualization

Background:

• High-dimensional data analysis is crucial in genomics.
• Dimension reduction techniques like PCA are common but have limitations.
• t-distributed stochastic neighbor embedding (t-SNE) is underutilized in human genetics.

Purpose of the Study:

• To evaluate the applicability of t-SNE for human genetic data.
• To compare t-SNE with principal component analysis (PCA) for visualizing population structure.
• To assess t-SNE's robustness and ability to reveal multi-scale genetic patterns.

Main Methods:

• Application of t-distributed stochastic neighbor embedding (t-SNE) to human genetic datasets.
• Comparison of t-SNE with principal component analysis (PCA) for dimension reduction and visualization.
• Analysis of t-SNE's performance in the presence of outliers and its capacity for multi-scale pattern detection.

Main Results:

• t-SNE successfully separates human samples by continental origin, similar to PCA.
• t-SNE demonstrates greater robustness to outliers compared to PCA.
• t-SNE effectively visualizes both continental and sub-continental population structures simultaneously.

Conclusions:

• t-SNE is a valuable tool for visualizing population stratification in human genetic data.
• The multi-scale pattern detection capability of t-SNE can enhance human genetic association studies.
• t-SNE offers a robust alternative or complement to PCA for genetic data analysis.