OT-scOmics

OT-scOmics leverages Optimal Transport to compute cell-cell similarity metrics from single-cell omics molecular profiles for improved similarity inference and unsupervised clustering.

Key Features:

• Optimal Transport as a Similarity Metric: Employs Optimal Transport to define distances between cells by comparing high-dimensional molecular profiles represented as probability distributions.
• Entropic Regularization: Incorporates entropic regularization into the OT distance calculation to accelerate computations while preserving accuracy.
• Comprehensive Benchmarking: Benchmarked against state-of-the-art metrics across thirteen independent datasets, including simulated data and single-cell omics datasets such as scRNA-seq, scATAC-seq, and single-cell DNA methylation.
• Improved Clustering Performance: Demonstrates superior cell-cell similarity inference and unsupervised clustering performance on real and simulated scRNA-seq datasets, with comparable performance to Pearson correlation for scATAC-seq and single-cell DNA methylation data.

Scientific Applications:

• Cellular Heterogeneity Analysis: Enables analysis of cellular heterogeneity across developmental processes and disease states by providing more nuanced cell-cell similarity measures.
• Unsupervised Clustering: Improves input similarity for unsupervised clustering methods to aid discovery of novel cell types or states.

Methodology:

Computational steps include measuring cell similarity using Optimal Transport on molecular profiles; applying entropic regularization to the OT distance calculation to address computational demands; and conducting extensive benchmarking against state-of-the-art metrics across thirteen independent datasets.