Search papers, labs, and topics across Lattice.
This paper introduces DMT-Dens, a novel manifold visualization method that addresses the challenge of preserving sampling density in low-dimensional embeddings of high-dimensional biological data. By leveraging a latent-token Transformer encoder and optimizing a loss function based on Pearson correlation, DMT-Dens effectively maintains the density of cell-state populations while ensuring competitive label separability. Benchmark evaluations confirm its superior performance in density preservation compared to existing methods, enhancing the interpretability of complex biological datasets.
DMT-Dens achieves unprecedented density preservation in biological data visualizations, crucial for accurately interpreting rare and transitional cell states.
Motivation: Low-dimensional embeddings are widely used to explore cell-state heterogeneity in single-cell and other high-dimensional biological data. Although many methods preserve local neighborhoods, they may distort the apparent sampling density of processed observations, altering the visual contrast between dense and sparse regions and complicating the interpretation of rare, transitional, or continuous cell-state populations. Results: We present DMT-Dens, a parametric manifold-visualization method built on a latent-token Transformer encoder. The model integrates rank-based manifold alignment with hard-pair aggregation. To preserve density, it optimizes a loss based on the Pearson correlation between k-nearest-neighbor log-radius estimates in the processed input and two-dimensional embedding spaces. Benchmark evaluations demonstrate strong density preservation, particularly on biological datasets, while retaining competitive label separability. Availability: Source code, data-processing scripts, and resolved experiment configurations are available at https://github.com/Ruizhe-wang/DMT-Dens.
