scTranslation: A Comprehensive Benchmark for Single-Cell Multi-Omics Modality Translation

2026-06-02

AI Summary

This study addresses the lack of a systematic benchmark for single-cell multi-omics modality translation, which has hindered method comparison and progress. To bridge this gap, the authors present the first comprehensive benchmarking platform, integrating diverse datasets, state-of-the-art translation models, and a unified preprocessing pipeline, along with a multidimensional evaluation framework. Through large-scale experiments, they systematically investigate how key factors—such as feature selection, data quality, and limited sample sizes—affect translation performance, thereby elucidating the strengths, limitations, and optimal use cases of current methods. The work further offers actionable insights to guide future algorithm development and provides an open-source, reproducible framework to foster standardized research in the field.

Abstract

Simultaneous measurement of multiple omics modalities in single cells enables researchers to gain a more comprehensive understanding of cellular states and regulatory mechanisms. However, due to high experimental costs, significant noise, and incomplete modality coverage, a variety of computational methods for modality translation have emerged in recent years. Despite the development of translation models, there is still a lack of systematic benchmark evaluation in terms of datasets, evaluation metrics, and influencing factors. To address this, we present scTranslation, a comprehensive benchmark for single-cell multi-omics modality translation tasks. It includes diverse translation datasets, integrates state-of-the-art models, and provides a comprehensive evaluation metrics. In addition, we assess model performance under different scenarios, such as feature selection, feature quality, and few-shot settings. These factors significantly affect model performance but have rarely been systematically studied before. Leveraging this benchmark, we conduct a large-scale study of current methods, report many insightful findings that open up new possibilities for future development. The benchmark is open-sourced to facilitate future research. The code is anonymously released at https://github.com/Bunnybeibei/scTranslation.

Problem

• single-cell multi-omics
• modality translation
• benchmark evaluation
• computational methods
• systematic assessment

Innovation

• single-cell multi-omics
• modality translation
• benchmarking
• few-shot learning
• feature selection