Sprecher
Beschreibung
In sequencing-based microscopy, spatial information is encoded through proximity networks, where edges represent molecular interactions. While intuitively, increased connectivity, termed here as network promiscuity, might be expected to degrade spatial reconstructions due to the presence of false edges, this study systematically examines whether promiscuity can, conversely, enhance reconstruction accuracy. By using synthetic random geometric graphs (RGGs) and experimental DNA microscopy datasets, we demonstrate that moderate network promiscuity substantially improves spatial reconstruction quality, even in the presence of noise. Our findings show that adding edges, contrary to conventional expectations, can increase true spatial signals sufficiently to offset the negative impact of false edges.
However, this benefit exists only up to critical connectivity thresholds (edge densities around 0.75 to 0.8), beyond which accuracy sharply declines due to excessive edge density limiting spatial resolution. Comparative analysis of reconstruction algorithms reveals distinct performance regimes: random-walk-based methods (STRND) do well in sparse, moderate-noise conditions, whereas shortest-path-based methods (MDS) outperform in densely connected, noisy environments. Additionally, experiments on real sequencing-based microscopy data confirm that controlled promiscuity enhances reconstruction robustness, aligning closely with synthetic predictions. These findings suggest a fundamental behavior governed by edge density, highlighting the practical value of intentionally designing promiscuous networks to achieve high-quality spatial reconstructions in DNA barcode network imaging.
