HBMAP: Bayesian inference of neural circuits from DNAbarcoded projection mapping

Abstract

Decoding how the brain routes information through its precise, long-range wiring remains a central challenge in neuroscience. Barcode-based mapping of axonal projections allows brain-wide, high-throughput investigation of projections at single-neuron resolution offering a powerful solution. However, principled methods for statistical analysis of barcode count data that can detect common projection rules and effectively integrate datasets across subjects are lacking. To address these issues, we developed a model-based clustering approach through hierarchical Bayesian mixtures which we call hierarchical Bayesian mapping of axonal projections (HBMAP). We show that the inferred model accurately reflects the features of the data and allows simultaneous identification of projection patterns and characterization of uncertainty that accounts for subject variability. Our study presents the first Bayesian approach to barcode-based projection mapping, offering a general solution for this class of data.