Cell-level random splits leak group-owned answers in single-cell benchmarks
The paper identifies and quantifies leakage in single-cell benchmarks caused by group-owned answers and introduces a tool to assess this leakage before model training.
Author
University of California, IrvineORCID 0000-0002-0044-6403Unclaimed ORCID profile · follow to get notified of new preprints
8 papers
Recent·Most cited
The paper identifies and quantifies leakage in single-cell benchmarks caused by group-owned answers and introduces a tool to assess this leakage before model training.
Dongning Chen, Nolan Origer, Sha Sun, Timothy L. Downing
This study shows XIST modulates the adhesome network and cell stemness by altering adhesion pathways and gene expression in ovarian cancer cells and across tissues.
This paper develops a species-native representation for Xenopus laevis to compare regenerative competence across species and identifies a vector competence ruler that distinguishes regenerative from fibrotic repair in mice.
Maria De Lourdes Andrade Ludena, Jonathan V. Duong, Aqsa Motiwala…·Sha Sun
This study identifies robust Xist expression in male Peromyscus leucopus somatic cells, showing gene-selective repression rather than global X-chromosome inactivation.
Shibashis Paul, Sha Sun, Karmella Ann Haynes, Tian Hong
This paper develops a reaction-diffusion model to explain the spatial control of gene silencing during X chromosome inactivation, showing how feedback loops and chromatin modifications regulate the spread of silencing.
Hu Cang, Huaxun Fan, Sha Sun, Kai Zhang
DistortCorrect Stitcher improves microscopy image stitching by integrating optical distortion correction, enhancing alignment accuracy for spatial transcriptomics data.
Heather M. Karner, Chiu-Ho T. Webb, Sarah Carmona…·Sha Sun
The paper investigates the functional conservation of lncRNA JPX across species despite sequence and structural divergence, showing it can functionally complement Jpx in mouse cells.