Abstract
<title>Abstract</title> <p>For a long time, we assumed somatic cells only pass DNA vertically during division. Recent work by Ly et al. [1] and Wu et al. [2] showed this rule breaks under heavy genomic stress: cells physically send intact DNA fragments, extrachromosomal circular DNA (ecDNA) oncogenes, and nuclear pieces to neighboring cells through thin tunneling nanotubes. Standard single-cell bioinformatic tools miss these events completely. Graph clustering algorithms smooth rare transfer transcripts right into main stroma clusters as technical noise, while doublet detectors like Scrublet [3] or DoubletFinder [4] search for droplet co-encapsulation artifacts and rank recipient cells as normal singlets. We built Nanotransfer to solve this problem. Following standard single-cell matrix processing [5,6], Nanotransfer combines four distinct modalities, Genomic Mosaicism Detector (GMD), Spatial Correlation Analyzer (SCA), ecDNA Transfer Score (ETS), and Allele-Specific Transfer Detector (ASTD) into a Stouffer Z-score and Benjamini-Hochberg FDR pipeline (q < 0.05). On synthetic Y-chromosome co-cultures, we achieved 89.80% sensitivity, 95.65% precision, and a low 4.35% FDR. Evaluating 3,500 primary human blood cells gave zero false calls (0.00% FPR). In a clinical glioblastoma spatial dataset (10x Visium, 4,880 spots), Nanotransfer isolated two transfer spots (0.04% of total area) forming a tight micro-cluster at the invasive tumor-stroma border with significant spatial autocorrelation (p = 0.0320). Nanotransfer is open-source on PyPI (pip install nanotransfer) and available as an enterprise web platform (https://nanotransfer.io).</p>