ACSNI

ACSNI predicts tissue- and disease-specific pathway components from gene expression profiles (GEPs) to infer pathway activities and guide molecular mechanism discovery.

Key Features:

• Integration with Prior Biological Knowledge: Integrates prior biological knowledge about processes and pathways to enhance predictive relevance.
• Deep Neural Network Utilization: Uses a deep neural network to decompose gene expression profiles into multi-variable pathway activities.
• Unsupervised Learning Approach: Operates via unsupervised machine learning and does not require labeled training data.
• Identification of Unknown Pathway Components: Predicts previously unknown components of signaling pathways, exemplified by mTOR, ATF2, and HOTAIRM1.
• Facilitation of Molecular Mechanism Discovery: Enables inference of molecular mechanisms from predicted pathway activities and components.
• Gene Prioritization for Targeted Experiments: Prioritizes candidate genes within pathways to guide targeted experimental design.

Scientific Applications:

• Pathway Analysis: Dissects signaling pathways across tissues and diseases to identify component activities.
• Gene Expression Studies: Analyzes gene expression data to reveal regulatory interactions and pathway activity patterns.
• Disease Research: Identifies disease-specific pathway alterations that can suggest therapeutic targets or biomarkers.

Methodology:

Inputs gene expression profiles (GEPs) that an unsupervised deep neural network decomposes into multi-variable pathway activities, predicts known and unknown pathway components (e.g., mTOR, ATF2, HOTAIRM1), and validates predictions against genetic perturbation and transcription factor binding datasets.