GILEA: In silico phenome profiling and editing using GAN Inversion.

Area of Science:

• Biomedical data science
• Quantitative biology
• Medical imaging analysis

Background:

• Data-driven modeling of heterogeneous disease states offers significant potential for advancing biomedical research.
• Analyzing phenotypic heterogeneity in complex biomedical datasets and emerging imaging techniques presents challenges.

Purpose of the Study:

• To introduce a novel framework for quantitative and interpretable analysis of biomedical image data.
• To apply the framework for in silico phenome profiling and editing.

Main Methods:

• Development of the GAN Inversion-enabled Latent Eigenvalue Analysis (GILEA) framework.
• Application of GILEA to cellular imaging datasets using the Cell Painting protocol.

Main Results:

• Demonstration of GILEA's performance on multiplexed fluorescence imaging data.
• Biological validation through editing latent representations and simulating phenotype transitions.
• Quantitative results support dynamic phenotype transitions between physiological and pathological states.

Conclusions:

• GILEA provides a novel, broadly applicable approach for quantitative and interpretable analysis of biomedical images.
• The framework facilitates understanding of phenotypic heterogeneity and disease state transitions.
• GILEA code and video demonstrations are publicly available for broader research application.