Integrating Computational and Experimental Techniques to Decipher Neuronal Heterogeneity

Key Takeaways

• Single‑nucleus RNA‑seq avoids shape‑bias in brain cell profiling
• AI designs enhancers targeting chronic‑pain neuronal subtypes
• 10x Xenium spatial transcriptomics resolves enhancer specificity
• Cell‑type‑specific chemogenetics blocks chronic pain without affecting touch
• Automation aims to close loop between experiments and AI decisions

Summary

Andreas Pfenning’s lab at Carnegie Mellon is merging single‑nucleus RNA‑seq, ATAC‑seq and high‑resolution spatial transcriptomics to map neuronal and glial subtypes without the shape‑bias of traditional droplet methods. AI algorithms then design cell‑type‑specific enhancers, which are screened on the 10x Xenium platform to confirm precise targeting. The approach has yielded chemogenetic tools that silence chronic‑pain circuits in mice while preserving normal sensation. Future work aims to automate the experimental‑computational loop and apply the technology to disorders such as Parkinson’s disease.

Pulse Analysis

The brain’s cellular mosaic presents a formidable obstacle for researchers seeking to link specific neuron types to behavior and disease. Traditional single‑cell droplet platforms struggle with the diverse morphologies of neurons, leading to sampling bias. By shifting to single‑nucleus RNA‑sequencing and complementary ATAC‑sequencing, Pfenning’s team captures a more faithful transcriptomic and epigenomic snapshot across thousands of nuclei. Advanced statistical tools such as regression discontinuity help distinguish true discrete cell classes from continuous gradients, while spatial transcriptomics adds a geographic dimension that validates cell‑type boundaries in situ.

Artificial intelligence now extends beyond data crunching to experimental design. The lab trains models on single‑cell datasets to predict enhancer and promoter sequences that activate only in targeted subpopulations, such as the chronic‑pain neurons of the spinal cord. These AI‑generated regulatory elements are rapidly evaluated using 10x Xenium, which multiplexes barcoded enhancers and deconvolves their spatial expression patterns. Successful enhancers drive chemogenetic receptors, enabling researchers to silence pain‑related circuits with a systemic drug while leaving normal touch, movement, and breathing untouched—a proof‑of‑concept that showcases the therapeutic promise of cell‑type‑specific gene regulation.

Integrating computational foresight with wet‑lab execution signals a paradigm shift for neuroscience and drug development. Automation platforms in CMU’s labs can now feed AI predictions directly into liquid‑handling robots, creating a closed‑loop system that iteratively refines enhancer libraries based on real‑time experimental feedback. Coupled with the Vertebrate Genomes Project’s expanding comparative datasets, this pipeline offers unprecedented resolution for dissecting conserved and species‑specific neural mechanisms. As the community gathers at ABRF 2026, the exchange of best practices around spatial technologies and AI‑driven design will likely accelerate the translation of these methods from mouse models to human therapeutics.