DOT: A flexible multi-objective optimization framework for transferring features across single-cell and spatial omics

TL;DR

DOT is a flexible multi-objective optimization framework that integrates single-cell RNA sequencing with spatial omics data, improving cellular feature localization and gene expression estimation across various spatial resolutions.

Contribution

It introduces a novel, adaptable optimization model for transferring cellular features between single-cell and spatial omics data, accommodating different data types and prior knowledge.

Findings

Achieves state-of-the-art localization of cell features in spatial data.

Accurately estimates unmeasured gene expression in low-coverage spatial datasets.

Runs efficiently on standard computational resources.

Abstract

Single-cell RNA sequencing (scRNA-seq) and spatially-resolved imaging/sequencing technologies have revolutionized biomedical research. On one hand, scRNA-seq provides information about a large portion of the transcriptome for individual cells, but lacks the spatial context. On the other hand, spatially-resolved measurements come with a trade-off between resolution and gene coverage. Combining scRNA-seq with different spatially-resolved technologies can thus provide a more complete map of tissues with enhanced cellular resolution and gene coverage. Here, we propose DOT, a novel multi-objective optimization framework for transferring cellular features across these data modalities. DOT is flexible and can be used to infer categorical (cell type or cell state) or continuous features (gene expression) in different types of spatial omics. Our optimization model combines practical aspects related to tissue composition, technical effects, and integration of prior knowledge, thereby providing flexibility to combine scRNA-seq and both low- and high-resolution spatial data. Our fast implementation based on the Frank-Wolfe algorithm achieves state-of-the-art or improved performance in localizing cell features in high- and low-resolution spatial data and estimating the expression of unmeasured genes in low-coverage spatial data across different tissues. DOT is freely available and can be deployed efficiently without large computational resources; typical cases-studies can be run on a laptop, facilitating its use.