SelectML

SelectML automates genomic prediction model testing and comparison to evaluate genomic selection methods for crop breeding using genetic marker data.

Key Features:

• Model Diversity: Supports both linear mixed models and machine-learning-based approaches for genomic prediction.
• Model Comparison: Systematically compares the performance of linear mixed models versus machine-learning methods.
• Feature Selection and Data Transformation: Implements feature selection methods and data transformers that conform to the scikit-learn API.
• Automated Pipeline: Automates training, comparison, and selection of genomic selection models.
• Sampling Scenario Simulation: Uses simulated marker datasets with randomly-sampled (mixed) and unevenly-sampled (unbalanced) populations.
• Sensitivity Analysis: Assesses sensitivity of machine-learning methods to sampling bias relative to linear mixed models.

Scientific Applications:

• Genomic Prediction in Agriculture: Analyzes genetic marker data to inform crop breeding and genomic selection strategies.
• Model Performance Evaluation: Evaluates relative performance of genomic selection methods under mixed and unbalanced population sampling scenarios.

Methodology:

An automated pipeline systematically tests and compares linear mixed models against machine-learning-based approaches using simulated marker datasets that include randomly-sampled (mixed) and unevenly-sampled (unbalanced) populations to quantify performance and sensitivity to sampling bias.