Skip to content
DraconiticResearch
Search Sources API
206,690 studies31 sourcesUpdated 20 min ago
Peer-reviewedCC BY 4.0Mathematics

Benchmarking Test Statistics and False Discovery Rate Procedures for Single-Cell Differential Expression

  • Bowen Duan1,2
  • Qingmin Zhang1
  • Peibo Hao2,3
  1. 1Yunnan Key Laboratory of Statistical Modeling and Data Analysis, Yunnan University, Kunming 650500, China
  2. 2National Center for Mathematics and Interdisciplinary Sciences, Academy of Mathematics and Systems Science, Chinese Academy of Sciences, Beijing 100190, China
  3. 3School of Mathematical Sciences, University of Chinese Academy of Sciences, Beijing 100049, China
Open at sourceDOICite

Abstract

Single-cell differential expression analysis combines a gene-level statistic with a false discovery rate (FDR) procedure. We examine how these two choices affect error control and power by pairing four statistics with BH, e-BH, data-splitting mirror, and target–decoy procedures. The benchmark uses zero-inflated negative-binomial simulations and a donor-level model fitted to Kang18 control data, with known signal genes in both designs. We vary dropout, signal type, gene number, and cross-gene dependence. Higher dropout reduces power across all 16 combinations in the mean-shift experiment, whereas null FDR depends on the statistic and its tail calibration. Distribution-sensitive statistics detect equal-mean dispersion changes that mean-based statistics largely miss. The apparent power gain of data splitting and target–decoy at a larger gene count is not reproduced when the number of signals is held fixed. The working e-value is sensitive to null-score calibration, and donor dependence affects some procedures without a common monotone pattern. These results show why benchmark comparisons should separate the signal being tested, the calibration of the input evidence, and the multiple-testing rule.

Related

  1. Reproducibility of social science research using aggregate statistics with noise infused for differential privacy
    Ryan Steed, Eduardo Abraham Schnadower Mustri, Alessandro Acquisti
    Proceedings of the National Academy of Sciences of the United States of AmericaPeer-reviewedCC BY-NC-NDFull textMolecular biology & gene regulationEurope PMC
  2. Benchmarking graph-based models for in-silico toxicity prediction in drug discovery
    Noel Suarez-Barro, Manuel Lama, Juan C. Vidal
    arXivPreprintarXiv licenceFull textBiophysicsBioinformatics & computational biologyarXiv
  3. Benchmarking AI scientists for omics data–driven biological discovery
    Erpai Luo, Jinmeng Jia, Yifan Xiong +6
    BioinformaticsPeer-reviewedCC BY25 citedMolecular biology & gene regulationOpenAlex
  4. Assessing the impact of parental linear gene normalization on the performance of statistical models for circular RNA differential expression analysis
    Briefings in bioinformaticsPeer-reviewedCC BY-NCFull textMolecular biology & gene regulationEurope PMC
  5. Interlaboratory benchmarking of differential scanning fluorimetry reveals sources of variability in protein thermal stability parameters
    S. Brule, M. Chevreuil, O. Burastero +40
    bioRxivPreprintCC BYFull textBiophysicsbioRxiv

Cite

BibTeXRIS
BibTeX
@article{duan2026benchmarking,
  title = {{Benchmarking Test Statistics and False Discovery Rate Procedures for Single-Cell Differential Expression}},
  author = {Bowen Duan and Qingmin Zhang and Peibo Hao},
  journal = {Mathematics},
  year = {2026},
  month = oct,
  doi = {10.3390/math14203786},
  url = {https://www.mdpi.com/2227-7390/14/20/3786},
  note = {CC BY 4.0},
}
RIS
TY  - JOUR
TI  - Benchmarking Test Statistics and False Discovery Rate Procedures for Single-Cell Differential Expression
AU  - Bowen Duan
AU  - Qingmin Zhang
AU  - Peibo Hao
PY  - 2026
DA  - 2026/10/19
JO  - Mathematics
DO  - 10.3390/math14203786
UR  - https://www.mdpi.com/2227-7390/14/20/3786
AB  - Single-cell differential expression analysis combines a gene-level statistic with a false discovery rate (FDR) procedure. We examine how these two choices affect error control and power by pairing four statistics with BH, e-BH, data-splitting mirror, and target–decoy procedures. The benchmark uses zero-inflated negative-binomial simulations and a donor-level model fitted to Kang18 control data, with known signal genes in both designs. We vary dropout, signal type, gene number, and cross-gene dependence. Higher dropout reduces power across all 16 combinations in the mean-shift experiment, whereas null FDR depends on the statistic and its tail calibration. Distribution-sensitive statistics detect equal-mean dispersion changes that mean-based statistics largely miss. The apparent power gain of data splitting and target–decoy at a larger gene count is not reproduced when the number of signals is held fixed. The working e-value is sensitive to null-score calibration, and donor dependence affects some procedures without a common monotone pattern. These results show why benchmark comparisons should separate the signal being tested, the calibration of the input evidence, and the multiple-testing rule.
SN  - 2227-7390
ER  -