Current Work

Whole-brain imaging in behaving fruit flies
With Mala Murthy and Thomas Clandinin

How do animals combine information from multiple sensory modalities, and how do these sensory representations shape behavior? I am broadly interested in characterizing whole-brain representations of sensory stimuli, and how these representations modulate an animal’s behavior. We have developed methods of two-photon functional imaging and audio-visual stimulus presentations, and are performing whole-brain imaging in behaving animals to study brainwide multi-sensory integration in the fruit fly Drosophila melanogaster. We have also developed a suite of software tools capable of precisely aligning in vivo whole-brain volumes to EM connectomes with ~5 micron precision, enabling high-resolution function-structure comparisons.

Relevant papers: Brezovec*, Lin*, et al. (PNAS 2024).

Analyzing the whole-brain fly connectome
With Mala Murthy, Sebastian Seung and Flywire Consortium

The FlyWire project is a large-scale collaborative connectomics proofreading effort, with contributions hundreds of scientists across many labs. With 140,000 neurons and millions of synapses, this connectome is currently the largest biological neuronal network to be densely reconstructed. I examined the network statistics of the fly’s wiring diagram, quantified topological properties, and identified populations of highly connected neurons. I also mapped mesoscale connectivity between 78 anatomically distinct brain regions, uncovering long-range directed and reciprocal connections. I found that despite its low connection probability, the Drosophila brain was highly non-random in its topology. Examining the frequency at which two-neuron and three-neuron motifs occur in the brain demonstrated its highly recurrent nature. I also mapped the topological distances of neurons in the brain from each sensory input (auditory, visual, chemosensory, etc.). Together, these findings provide a framework for future experimental and theoretical work in Drosophila neuroscience.

Relevant papers: Dorkenwald et al. (Nature 2024), Lin et al. (Nature 2024).

Prior Work

Modeling pan-neuronal activity using a constrained neural network
With Lu Mi and Srinivas Turaga

Could a modeling-based approach to the relationship between functional correlations in pan-neuronal data and the connectome yield more insight? Working with Dr. Srinivas Turaga’s group, we constructed an encoder-decoder neural network where edges were defined by the C. elegans connectome and weights were trained based on labeled pan-neuronal data. This model has demonstrated improved accuracy when predicting the activity of downstream neurons.

Relevant papers: Mi et al. (ICLR 2022).

Deterministic multicolor pan-neuronal landmark imaging
With Eviatar Yemini, Aravinthan Samuel, Vivek Venkatachalam, Oliver Hobert and Liam Paninski

In collaboration with Dr. Eviatar Yemini in Dr. Oliver Hobert’s lab, we developed worms which have a stereotyped multicolor fluorescence map (NeuroPAL) which allows for the comprehensive identification of all 300 neurons in the C. elegans nervous system. Such a map allows us to capture labeled pan-neuronal activity in C. elegans for the first time, giving us the ability to directly compare experiments, average data across animals, and interpret activity in the context of the connectome. We optimized the strain for live imaging, developed experimental methods for acquiring high-resolution multicolor landmark volumes, and performed chemosensory experiments with pan-neuronally labeled animals. In collaboration with Dr. Liam Paninski’s group, we developed software to semi-automatically ID neurons and demix fluorescence signals of neighboring neurons. We found that a large fraction of neurons are engaged by even simple stimuli, and these activity patterns were distinct for different stimuli. We also found little to no correlation between functional activity and synaptic weights in the C. elegans connectome.

Relevant papers: Yemeni, Lin et al. (Cell 2021), Nejatbakhsh et al. (MICCAI 2020).

Ensemble representations of olfactory stimuli in C. elegans
With Aravinthan Samuel, Vivek Venkatachalam, Mei Zhen and Cengiz Pehlevan

Despite having only 11 pairs of chemosensory neurons, C. elegans is capable of detecting and discriminating a wide range of odorants. We know from the C. elegans connectome that many of these neurons are wired to each other, and some also receive feedback from interneurons. In collaboration with Dr. Mei Zhen’s lab, we generated new C. elegans lines in which the entire chemosensory ensembled is labeled with GCaMP. We designed and built microfluidics devices to deliver odorants with high temporal precision. Using these devices, we presented animals with a broad range of odorants spanning many chemical families while simultaneously recording calcium activity using a spinning-disk confocal microscope. From these data, we built a map of odor representation in the sensory ensemble, uncovering previously unreported responses to chemosensory stimuli.  Working with Dr. Cengiz Pehlevan’s group, we built classifiers which demonstrated theoretical discriminability between odors.

Relevant papers: Lin et al. (Science Adv. 2023).

in vivo quantification of mRNA transcription in Drosophila embryos
With Thomas Gregor and Hernan Garcia

My undergraduate research in Dr. Thomas Gregor’s lab focused on developing a method for visualizing the loci of mRNA transcription, using an MCP-MS2 stem loop system driven by a gap gene promoter of interest to tag mRNA with fluorescent markers during transcription. We generated new fly lines and quantified the production of mRNA in the early Drosophila embryo. From these data, we were able to extract nucleus-level parameters of RNA polymerase activity. We also developed a dual-reporter experiment, quantifying the transcriptional noise with nucleus-width spatial resolution and proposing biological models for the intrinsic and correlated noise components.

Relevant papers: Garcia et al. (Curr. Bio. 2013).