Researchers have developed a new method called DMT-Dens for visualizing high-dimensional biological data. The goal is to preserve the density of the data points in lower-dimensional representations, which can help identify rare or transitional cell states. This is achieved by optimizing a loss function that correlates the log-radius estimates of nearest neighbors between the original and embedded spaces. The method has been benchmarked on various biological datasets and shows strong performance in preserving density while maintaining label separability. --- Why it matters: This matters to researchers working with single-cell or high-dimensional biological data because it provides a more accurate visualization tool that preserves the underlying structure of the data, making it easier to identify rare cell states and interpret results. Source: https://arxiv.org/abs/2608.17571