FIND

FIND predicts genome-wide spatio-temporal gene expression during Drosophila embryonic development.

Key Features:

• Machine Learning-Driven Predictions: FIND employs machine learning to generate quantitative predictions of gene expression across more than 200 tissue-developmental stages.
• Training Data and Integration: Models are trained on a compendium of over 6,378 genome-wide expression and chromatin profiling experiments and integrate signals in a cell lineage-aware manner.
• Structured In Silico Nano-Dissection: Implements a structured in silico nano-dissection approach to resolve tissue- and stage-specific expression patterns.
• Validation and Accuracy: Predictions were evaluated by cross-validation and 22 novel predictions were experimentally confirmed across four embryonic tissues.
• Exploratory Query Functions: Provides gene query (single or bulk), tissue-stage queries, similar-genes identification, and tissue-finder functions for analysis of gene lists and spatiotemporal expression profiles.

Scientific Applications:

• Enhancing in situ hybridization data: Supplies quantitative spatiotemporal expression predictions to complement qualitative in situ hybridization annotations, including non-coding genes.
• Extracting tissue specificity signals: Enables derivation of tissue-specific signals from non-tissue-dissected genome-wide experiments.
• Disease modeling prioritization: Assists prioritization of tissues and developmental stages for disease modeling based on predicted expression patterns.

Methodology:

Machine learning models trained on a compendium of >6,378 genome-wide expression and chromatin profiling experiments, integrating expression and chromatin signals in a cell lineage-aware manner, applying a structured in silico nano-dissection approach and evaluated by cross-validation.