GLARE: discovering hidden patterns in spaceflight transcriptome using representation learning.

Area of Science:

• Space Biology
• Transcriptomics
• Machine Learning

Background:

• Spaceflight exposes organisms to unique stressors, offering insights into biological processes.
• NASA's GeneLab provides public access to numerous spaceflight transcriptomic datasets.
• Machine learning approaches have not been broadly applied to spaceflight omics data.

Purpose of the Study:

• To develop an open-source pipeline (GLARE) for analyzing spaceflight transcriptomic data using representation learning.
• To enhance the performance of downstream analytical tasks on spaceflight datasets.
• To uncover novel biological insights from existing spaceflight data.

Main Methods:

• Developed GLARE (GeneLab Representation learning pipelinE), an open-source pipeline.
• Trained various representation learning models, including manifold and self-supervised learning.
• Applied GLARE to the CARA spaceflight experiment's Arabidopsis root tip transcriptome data (light, dark, microgravity).

Main Results:

• GLARE substantiated previous findings on cell wall remodeling in Arabidopsis under spaceflight conditions.
• GLARE identified additional gene expression patterns related to spaceflight treatments.
• Evidence of hypoxic response was revealed by GLARE analysis.

Conclusions:

• GLARE effectively analyzes spaceflight transcriptomic data, revealing new biological patterns.
• Machine learning-driven analysis of spaceflight omics data holds significant potential.
• GLARE can supplement initial findings from spaceflight biology studies.