2020 connectomics

A connectome and analysis of the adult Drosophila central brain

Scheffer, Xu, Januszewski et al.

· DOI

Opportunity

Despite the centrality of neural circuits to understanding animal behavior, comprehensive wiring diagrams of even small brains remained elusive due to the enormous scale of imaging and reconstruction required.

Challenge

Mapping an entire adult Drosophila central brain required petabyte-scale electron microscopy data, new alignment and segmentation algorithms, synapse detection across billions of pixels, and extensive proofreading to achieve reliable connectivity.

Action

The authors developed and applied improved procedures for sample preparation, FIB-SEM and transmission EM imaging, image alignment, automated segmentation, synapse detection, and collaborative proofreading to reconstruct the adult fly central brain. They defined cell types, refined compartment boundaries, and built an exhaustive atlas of neurons and their synaptic connections.

Resolution

The resulting connectome covers the large majority of the Drosophila central brain with detailed chemical-synapse connectivity, novel cell types, and an openly accessible dataset. Analyses revealed distributions of connection strengths, neural motifs at multiple scales, electrical implications of compartmentalization, and evidence that packing density is an evolutionary constraint.

Future Work

Functional validation of the identified circuits and extension of the approach to other species or brain regions with larger neuron counts remain key open challenges.

RNA sequencing (RNA-Seq) reads out all the RNA molecules present in a cell, revealing which genes are active and to what degree. This foundational paper described and demonstrated the method, showing it could detect far more than older gene chip methods. It became the basis for understanding what makes different brain cell types molecularly distinct.

The Mortazavi et al. RNA-Seq paper demonstrated high-coverage paired-end sequencing of the mouse transcriptome, introducing key computational concepts including reads per kilobase per million mapped reads (RPKM) for expression normalization and splice junction mapping for isoform detection. In neuroscience, RNA-Seq enabled systematic characterization of neuron type-specific transcriptomes, forming the molecular basis of cell-type atlases used to interpret connectomics data.

This paper established many foundational computational practices for RNA-Seq analysis: read alignment to spliced reference genomes, RPKM normalization, and differential expression analysis. Key limitations acknowledged included PCR amplification bias, 3' bias from poly-A selection, and difficulties with very long or very short transcripts. In connectomics, RNA-Seq-derived transcriptomic cell type classifications (as in the Allen Brain Cell Atlas) provide molecular identity for neurons that can be matched to morphological and connectivity-based classifications from EM data, enabling multimodal cell type reconciliation.

Paper

drosophila fib sem whole brain neuprint proofreading deep learning