This project is a collaboration across two universities and multiple scientific disciplines to develop new scalable
3D molecular imaging and analysis approaches for cell type identification within human brain tissue. We will
focus our efforts on the olfactory system, comprising the olfactory epithelium (OE) and the olfactory bulb (OB).
This system is an ideally confined human model system to build and test a new suite of scalable tools for the
generation of a human brain atlas because the connectomics of olfactory sensory neurons to the bulb is dictated
by olfactory receptor expression. Our long term goals are to provide the community with new “physics-first”
methods that improve scalability, rigor, and 3D measurements for cell census creation and create the first spatial
map of connections between the OE and OB. We plan to achieve our goals across two specific aims, carried out
in parallel. In our first aim, we will scale up high-resolution, high-speed single objective light-sheet microscopy
for 3D imaging of proteins and RNA in the human olfactory bulb. Using linear unmixing, we will image up to 8
proteins. Using iterative amplified RNA-FISH labeling by fluidic exchange, we will initially image 130 RNAs and
detail plans to expand the number of RNA. In our second aim, we will develop a Bayesian nonparametric image
analysis framework that self-consistently and simultaneously determines the probability associated with all RNA
locations, numbers, and identities in the presence of variable autofluorescence and variable readout efficiency.
Within the Bayesian paradigm, we propose a new approach to error correction in barcoded fluorescence
experiments that significantly reduces the number of rounds required. We will apply these combined fundamental
improvements to map full sections of the OB to determine the targeting of glomeruli by olfactory sensory neurons
expressing specific olfactory receptors. As olfactory receptors have high homology and sparse expression, both
in situ hybridization and in situ sequencing may report a high number of false positives or false negatives. To
mitigate potential identification errors, we will evaluate dual-color amplified labeling strategies combined with
Bayesian nonparametric analysis that assigns probabilities based on a self-consistent analysis of all image
stacks across all colors simultaneously to avoid drawing conclusions based on local assessments of the identity
of a bright spot. Combining these methodological advancements, we will generate 3D spatial maps of confidence
intervals for cell types and individual olfactory receptors expression across the olfactory system.