SCOT: Single-Cell Multi-Omics Alignment with Optimal Transport.

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Single-cell sequencing technologies enable detailed genomic analysis.
• Simultaneously applying multiple sequencing assays to the same single cell is limited.
• Computational integration of multi-omic data is essential for joint single-cell analyses.

Purpose of the Study:

• To develop an unsupervised algorithm for aligning single-cell multi-omics datasets.
• To address the challenge of integrating data without sample-wise or feature-wise correspondences.
• To improve the accuracy and efficiency of single-cell data integration.

Main Methods:

• Developed single-cell alignment with optimal transport (SCOT), an unsupervised algorithm.
• Utilized Gromov-Wasserstein optimal transport for aligning single-cell multi-omics datasets.
• Implemented a self-tuning heuristic for hyperparameter selection based on Gromov-Wasserstein distance.

Main Results:

• SCOT performs comparably to state-of-the-art unsupervised alignment methods.
• SCOT demonstrates improved speed and requires fewer hyperparameter tunings.
• SCOT achieves superior alignment in the unsupervised setting without orthogonal correspondence information.

Conclusions:

• SCOT offers an effective and efficient solution for integrating single-cell multi-omics data.
• The self-tuning heuristic enhances SCOT's performance in unsupervised alignment.
• SCOT advances the capability for joint analyses of diverse single-cell omics datasets.