A Hybrid Supervised Approach to Human Population Identification Using Genomics Data.

Area of Science:

• Genetics and Genomics
• Bioinformatics
• Population Genetics

Background:

• Single nucleotide polymorphisms (SNPs) are key genetic variations used to trace an individual's ancestry.
• Understanding population structure is crucial for personalized medicine and disease association studies.
• Principal Component Analysis (PCA) is a common unsupervised method for population structure analysis.

Purpose of the Study:

• To investigate supervised machine learning approaches for predicting an individual's genetic origin.
• To introduce and evaluate a novel hybrid method, PCA-LASSO, combining PCA and LASSO regression.
• To assess the accuracy of PCA-LASSO in identifying population of origin using SNP data.

Main Methods:

• Exploration of supervised classification methods, specifically LASSO multinomial regression.
• Development of PCA-LASSO, integrating Principal Component Analysis with LASSO regression.
• Application and validation of PCA-LASSO on the 1,000 Genomes Project dataset.

Main Results:

• LASSO multinomial regression demonstrated effectiveness in classifying individuals' genetic origin.
• The proposed PCA-LASSO method significantly improved prediction accuracy compared to existing approaches.
• PCA-LASSO achieved high accuracy in predicting individual's origin population on a large-scale dataset.

Conclusions:

• PCA-LASSO offers a powerful and accurate tool for population origin inference.
• This method has significant implications for population genetics research and personalized medicine.
• The integration of PCA and LASSO provides a robust framework for genetic ancestry analysis.