Exploring genomic feature selection: A comparative analysis of GWAS and machine learning algorithms in a large-scale...

Area of Science:

• Genomics
• Bioinformatics
• Plant Breeding

Background:

• High-throughput sequencing yields vast genomic datasets, necessitating advanced methods for genetic marker discovery.
• The
• small n large p
• challenge (few samples, many features) complicates identifying relevant genetic markers for complex traits.

Purpose of the Study:

• To evaluate and compare the effectiveness of different feature selection methodologies in genomic data analysis.
• To identify predictive genetic markers for complex traits in soybean using machine learning and traditional approaches.
• To assess the impact of feature selection on predictive modeling accuracy for various phenotypes.

Main Methods:

• Utilized a large soybean (Glycine max L. Merr.) dataset with 966 lines and over 5.5 million single nucleotide polymorphisms.
• Compared traditional genome-wide association studies (GWAS) with machine learning algorithms: random forest and extreme gradient boosting.
• Assessed feature selection performance by constructing predictive models and evaluating prediction accuracies for different phenotypes.

Main Results:

• Machine learning methods (random forest, extreme gradient boosting) demonstrated strong performance in pinpointing predictive genetic features.
• Comparative analysis revealed varying strengths and limitations of each method in handling high-dimensional genomic data.
• Selected features optimized predictive models, indicating the utility of these approaches for marker discovery.

Conclusions:

• Feature selection is crucial for enhancing interpretability and computational efficiency in large-scale genomic studies.
• Random forest and extreme gradient boosting offer powerful alternatives to traditional GWAS for identifying significant genetic markers.
• The study provides insights into optimizing genomic data analysis for trait prediction and marker discovery in soybean breeding programs.