Connections, Cutoffs and Coincidences: Increasing Throughput and Reliability in Barcoded Rabies Connectome Inference using Computational Analysis

Advances in viral barcoding techniques offer an unprecedented opportunity to obtain connectomes at scale through infection tracing with snRNASeq. However, standard methods for analyzing this new modality relies on hard thresholds that discard a large fraction of observations and inferable synapses. Synapses are then often inferred based on individual barcodes unique to a single spreader cell, which can lead to false positives due to coincidences in primary or secondary co-infections, particularly in dual-helper protocols with non-uniform barcode libraries. We discuss how noise & signal looks like in the data, analyze how they impact interpretation of results with standard analysis methods, and propose a computational method using graph set transformers that allow use of 100% of the observations, including ambient droplets, increasing both throughput and reliability of inferred synapses.

Learning Objectives:

1. Describe the experimental principles of viral barcoding with snRNASeq and how it can be utilized to trace infection paths for connectome mapping.

2. Differentiate between true biological signal and technical noise to identify causes of false positives & negatives in standard threshold-based analysis.

3. Contrast proposed computational methods against traditional analysis techniques in terms of data utilization, throughput, and the reliability of inferred synapses.