The STP Open Challenge: A Community Effort to Define and Benchmark Spatial Transcriptome-to-Proteome Prediction

While spatial transcriptomics technologies are increasingly accessible, spatial proteomics remains constrained by high costs and limited throughput. Furthermore, poor mRNA-protein correlation, driven by post-transcriptional regulation, necessitates multi-omics integration for complementary biological insights. Current RNA-to-protein prediction models, primarily trained on bulk or single-cell data, often overlook critical spatial context. Consequently, it remains unclear which model architecture, spatial representation, or histopathological features best facilitate spatial protein abundance inference. To address these gaps, we defined the STP (Spatial Transcriptome-to-Proteome) task and launched the STP Open Challenge, a competition for predicting spatial proteomics from spatial transcriptomics of adjacent tissue sections. Utilizing a curated paired spatial dataset and a standardized evaluation framework, the challenge attracted 101 teams and yielded 896 valid submissions. Top-performing methods achieved strong predictive performance while preserving spatial organization. The landscape of submitted models revealed diverse modeling strategies and motivated us to develop a consensus-based ensemble framework that leverages various participant methods to enhance prediction robustness. Extended evaluation on unseen datasets demonstrated that the top-performing models outperformed existing baselines, highlighting the effectiveness of the community effort for this task. By formally defining the STP task and providing a standardized benchmarking platform, this work establishes a foundation for scalable, cost-effective spatial proteomics prediction and future multi-omics integration.

Authors

Institutions

Publication Details

Journal
Research Square
Published
2026-11-19
DOI
https://doi.org/10.21203/rs.3.rs-10262522/v1
Primary Topic
Single-cell and spatial transcriptomics
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

The STP Open Challenge: A Community Effort to Define and Benchmark Spatial Transcriptome-to-Proteome Prediction

Zhi‐Ping Liu, Hongxin Wei, Shimin Shuai, Yimeng Qiao et al.
Research Square
Single-cell and spatial transcriptomics
preprint

The STP Open Challenge: A Community Effort to Define and Benchmark Spatial Transcriptome-to-Proteome Prediction

Zhi‐Ping Liu, Hongxin Wei, Shimin Shuai, Yimeng Qiao, Gang Hu, Fuhao Guo, Ran Liu, Jiguang Wang, Yongbing Zhang, L. Xu, Yuyan Ruan, Jiancheng Zhong, Mingyue Wei, Zhidong Yang, Zijie FANG, Jihong Tang, Kai Zhang, Yixuan Li, Lang Wu
preprint en

Abstract

While spatial transcriptomics technologies are increasingly accessible, spatial proteomics remains constrained by high costs and limited throughput. Furthermore, poor mRNA-protein correlation, driven by post-transcriptional regulation, necessitates multi-omics integration for complementary biological insights. Current RNA-to-protein prediction models, primarily trained on bulk or single-cell data, often overlook critical spatial context. Consequently, it remains unclear which model architecture, spatial representation, or histopathological features best facilitate spatial protein abundance inference. To address these gaps, we defined the STP (Spatial Transcriptome-to-Proteome) task and launched the STP Open Challenge, a competition for predicting spatial proteomics from spatial transcriptomics of adjacent tissue sections. Utilizing a curated paired spatial dataset and a standardized evaluation framework, the challenge attracted 101 teams and yielded 896 valid submissions. Top-performing methods achieved strong predictive performance while preserving spatial organization. The landscape of submitted models revealed diverse modeling strategies and motivated us to develop a consensus-based ensemble framework that leverages various participant methods to enhance prediction robustness. Extended evaluation on unseen datasets demonstrated that the top-performing models outperformed existing baselines, highlighting the effectiveness of the community effort for this task. By formally defining the STP task and providing a standardized benchmarking platform, this work establishes a foundation for scalable, cost-effective spatial proteomics prediction and future multi-omics integration.

Research Square
Shandong University (CN), Hunan Normal University (CN), Hong Kong University of Science and Technology (HK), Harbin Institute of Technology (CN), Nankai University (CN), Southern University of Science and Technology (CN), Tsinghua University (CN)
Single-cell and spatial transcriptomics
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon