EmbryoNet

EmbryoNet classifies embryonic phenotypes and links them to signaling pathway perturbations using a deep convolutional neural network trained on time-resolved zebrafish embryo images.

Key Features:

• Automated phenotyping: Automatically identifies signaling mutants in zebrafish embryos from imaging data with high accuracy.
• Deep convolutional neural network: Employs a CNN architecture trained on over 2 million zebrafish embryo images for robust phenotype classification.
• Temporal modeling of development: Incorporates time-dependent developmental trajectories to track and classify phenotypic defects across stages.
• Signaling pathway classification: Classifies phenotypic defects associated with loss of function in seven major evolutionarily conserved signaling pathways.
• Cross-species applicability: Detects signaling defects in evolutionarily distant species beyond zebrafish.
• Drug mechanism inference in screens: Infers mechanisms of action of pharmaceutical compounds by analyzing embryo phenotypic responses in high-throughput screens.

Scientific Applications:

• Developmental biology: Elucidates links between genetic perturbations and phenotypic outcomes in early vertebrate development.
• Drug discovery and toxicology: Resolves mechanisms of action and identifies developmental toxicants from high-throughput embryonic phenotyping.
• Comparative evolutionary studies: Enables cross-species comparison of conserved developmental signaling defects.

Methodology:

Deep convolutional neural network trained on over 2 million zebrafish embryo images with integration of time-dependent developmental trajectory modeling.