FuncISH: learning a functional representation of neural ISH images

Area of Science:

• Neuroscience
• Bioinformatics
• Computational Biology

Background:

• High-spatial resolution imaging of mammalian brains yields complex gene expression patterns.
• Analyzing these patterns requires extracting functionally relevant information and creating interpretable representations for neuroscientists.

Purpose of the Study:

• To develop a method (FuncISH) for learning functional representations of neural in situ hybridization (ISH) images.
• To enable meaningful interpretation of complex gene expression patterns by neuroscientists.

Main Methods:

• Representing ISH images using a histogram of local descriptors across multiple scales.
• Learning detectors for functional gene ontology (GO) categories for each image.
• Mapping images to a low-dimensional space where axes represent functional annotations.

Main Results:

• Achieved high accuracy in inferring neural biological processes from spatial expression patterns in mouse brain ISH data.
• Predicted gene interaction properties, including protein-protein interactions and cell-type specificity, more accurately than global correlation methods.
• Identified novel similarities in expression patterns of GABAergic neuronal markers and inferred new gene functions.

Conclusions:

• FuncISH provides a powerful tool for analyzing and interpreting large-scale neural ISH datasets.
• Functional representations enhance the understanding of gene function and interactions based on spatial expression data.
• The method facilitates discovery of novel biological insights from complex imaging data.