QOT: Efficient Computation of Sample Level Distance Matrix from Single-Cell Omics Data through Quantized Optimal...

Area of Science:

• Bioinformatics
• Computational Biology
• Data Science

Background:

• Single-cell technologies provide high-dimensional cell population data, posing computational challenges for comparative analyses.
• Generating sample-level distance matrices is crucial for understanding cellular architecture across diverse biological conditions.
• Optimal Transport (OT) is a geometric data analysis tool used in bioinformatics, but can be computationally intensive for large datasets.

Approach:

• We introduce Quantized Optimal Transport (QOT), an efficient method for computing sample-level distance matrices from large-scale single-cell omics data.
• QOT utilizes a quantization step to streamline the computation of distance matrices.
• The method is validated on real-world single-cell genomics and pathomics datasets.

Key Points:

• QOT significantly improves the accuracy and robustness of sample-level distance matrix generation compared to traditional OT algorithms.
• The generated distance matrices enable extrapolation of cell-level insights to sample-level categorizations.
• QOT effectively handles the complexity and volume of high-throughput single-cell omics data.

Conclusions:

• QOT offers a computationally efficient and accurate solution for analyzing large-scale single-cell data.
• The sample-level distance matrices derived from QOT are valuable for downstream analyses, such as uncovering disease progression trajectories.
• This work advances biomedical informatics and data science by providing a powerful tool for single-cell data analysis.