Back

Unlocking cross-modal interplay of single-cell and spatial joint profiling with CellMATE

Jingping, Y.; Qi, W.; Bolei, Z.; Luyu, G.; Yue, G.; Erguang, L.

2024-09-09 bioinformatics
10.1101/2024.09.06.610031 bioRxiv
Show abstract

A key advantage of single-cell multimodal joint profiling is the modality interplay, which is essential for deciphering the cell fate. However, while current analytical methods can leverage the additive benefits, they fall short to explore the synergistic insights of joint profiling, thereby diminishing the advantage of joint profiling. Here, we introduce CellMATE, a Multi-head Adversarial Training-based Early-integration approach specifically developed for multimodal joint profiling. CellMATE can capture both additive and synergistic benefits inherent in joint profiling through auto-learning of multimodal distributions and simultaneously represents all features into a unified latent space. Through extensive evaluation across diverse joint profiling scenarios, CellMATE demonstrated its superiority in ensuring utility of cross-modal properties, uncovering cellular heterogeneity and plasticity, and delineating differentiation trajectories. CellMATE uniquely unlocks the full potential of joint profiling to elucidate the dynamic nature of cells during critical processes as differentiation, development and diseases. Graphical abstracts O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=132 SRC="FIGDIR/small/610031v1_ufig1.gif" ALT="Figure 1"> View larger version (45K): org.highwire.dtl.DTLVardef@8a3f46org.highwire.dtl.DTLVardef@401d5aorg.highwire.dtl.DTLVardef@1472aecorg.highwire.dtl.DTLVardef@1530b62_HPS_FORMAT_FIGEXP M_FIG C_FIG

Matching journals

The top 4 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.