Prediction of a Cell-Class-Specific Mouse Mesoconnectome Using Gene Expression Data.

Area of Science:

• Computational Neuroscience
• Systems Neuroscience
• Genomics

Background:

• Reconstructing brain connectivity at high resolution for computational models is challenging.
• A mesoconnectome with laminar and cell-class specificity is needed for studying cognitive processes.

Purpose of the Study:

• To analyze gene expression patterns for predicting cell-class and layer-specific projection patterns.
• To assess functional annotations of predictive gene groups.
• To improve mesoconnectome reconstruction accuracy.

Main Methods:

• Used publicly available volumetric gene expression and connectivity data.
• Trained computational models (ridge regression) to predict axonal projections using gene expression.
• Identified spatial gene modules using dictionary learning and sparse coding.

Main Results:

• Ridge regression achieved a median r² of 0.54 for predicting projection strength.
• Binarized predictions yielded a median area under the ROC of 0.89.
• Spatial gene modules provided comparable prediction accuracy (median r² of 0.51).

Conclusions:

• Gene expression patterns can accurately predict cell-class and layer-specific brain connectivity.
• This predictive workflow enables multimodal data integration for improved mesoconnectome reconstruction.
• Findings support neuroscience use cases requiring detailed brain network information.