Comparing Deep Learning Approaches for Understanding Genotype × Phenotype Interactions in Biomass Sorghum

Area of Science:

• Agricultural Science
• Genetics
• Computer Science

Background:

• Understanding genotype-phenotype relationships is crucial for crop improvement.
• Biomass sorghum is a key crop for biofuel production.
• Deep learning offers novel approaches for analyzing complex biological data.

Purpose of the Study:

• To investigate the efficacy of deep convolutional neural networks (CNNs) in predicting genotype-phenotype relationships in biomass sorghum.
• To compare two CNN training strategies: direct SNP classification versus data-driven feature learning.
• To identify image regions critical for predicting genetic markers and uncover unknown genotype-phenotype associations.

Main Methods:

• Trained CNNs on overhead imagery of biomass sorghum.
• Employed two CNN approaches: direct SNP classification and data-driven feature learning for marker classification.
• Visualized image regions most influential for prediction accuracy.

Main Results:

• Both CNN approaches demonstrated efficiency in predicting the presence or absence of genetic markers.
• Data-driven feature learning CNNs achieved higher prediction performance.
• Visualizations for data-driven approaches were less interpretable than direct classification.

Conclusions:

• Deep learning, particularly data-driven CNNs, shows promise for predicting genotype-phenotype relationships in sorghum.
• Further research is needed to enhance the interpretability of data-driven models.
• This methodology could be applied to discover novel, unknown genotype-phenotype associations.