EUVIMEDEUVIMED entdecken
Crossref

GenAR: Next-scale autoregressive generation for spatial gene expression prediction

Jiarui Ouyang, Yihui Wang, Yihang Gao, Yingxue Xu, Shu Yang, Hao Chen

Medical Image Analysis · 2026-11 · 114 · 104232

DOI: 10.1016/j.media.2026.104232 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Medical Image Analysis, 2026-11. Verfasst von Jiarui Ouyang und 5 weiteren Autor:innen. Beteiligte Einrichtung: Department of Computer Science and Engineering, The Hong Kong University of Science and Technology, Hong Kong, China. Electronic address: jouyangag@c…. Hauptthemen laut MEDLINE: Gene Expression Profiling. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 43 Literaturangaben. Gefördert von: Shenzhen Science and Technology Innovation Committee, University Grants Committee, Research Grants Council, University Grants Committee.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Spatial Transcriptomics (ST) offers spatially resolved gene expression but remains costly. Predicting expression directly from widely available Hematoxylin and Eosin (H&E) stained images presents a cost-effective alternative. However, most computational approaches (i) predict each gene independently, overlooking co-expression structure, and (ii) cast the task as continuous regression despite expression being discrete counts. This mismatch can yield biologically implausible outputs and complicate downstream analyses. We introduce GenAR, a multi-scale autoregressive framework that refines predictions from coarse to fine. GenAR (a) clusters genes into hierarchical groups to expose cross-gene dependencies, (b) models expression as discrete token generation over a fixed vocabulary of integer count tokens to directly predict raw counts, and (c) conditions decoding on fused histological and spatial embeddings. By modeling expression on the physical count scale, GenAR avoids the limitations of continuous regression, while its coarse-to-fine factorization ensures a principled conditional decomposition. Extensive experimental results on five ST datasets across different tissue types demonstrate that GenAR achieves state-of-the-art performance, offering potential implications for precision medicine and cost-effective molecular profiling. Code is publicly available at https://github.com/oyjr/genar.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Gene Expression ProfilingSpatial TranscriptomicsAlgorithmsHumans

Stichwörter der Autor:innen

AutoregressiveGene expression predictionHistopathologySpatial transcriptomics

Autor:innen und Institutionen

  • Department of Computer Science and Engineering, The Hong Kong University of Science and Technology, Hong Kong, China. Electronic address: jouyangag@connect.ust.hk.
  • Department of Computer Science and Engineering, The Hong Kong University of Science and Technology, Hong Kong, China. Electronic address: ywangrm@connect.ust.hk.
  • Department of Computer Science and Engineering, The Hong Kong University of Science and Technology, Hong Kong, China; Shenzhen Loop Area Institute, Shenzhen, China. Electronic address: ygaodh@connect.ust.hk.
  • Department of Computer Science and Engineering, The Hong Kong University of Science and Technology, Hong Kong, China. Electronic address: yxueb@connect.ust.hk.
  • Department of Computer Science and Engineering, The Hong Kong University of Science and Technology, Hong Kong, China. Electronic address: syangcw@connect.ust.hk.
  • Department of Computer Science and Engineering, The Hong Kong University of Science and Technology, Hong Kong, China; Department of Chemical and Biological Engineering, The Hong Kong University of Science and Technology…

Förderung

  • Shenzhen Science and Technology Innovation Committee
  • University Grants Committee
  • Research Grants Council, University Grants Committee
  • Innovation and Technology Commission
  • National Natural Science Foundation of China

Zitieren

APA

Ouyang, J., Wang, Y., Gao, Y., Xu, Y., Yang, S., & Chen, H. (2026). GenAR: Next-scale autoregressive generation for spatial gene expression prediction. Medical Image Analysis, 114, 104232. https://doi.org/10.1016/j.media.2026.104232

Vancouver

Ouyang J, Wang Y, Gao Y, Xu Y, Yang S, Chen H. GenAR: Next-scale autoregressive generation for spatial gene expression prediction. Medical Image Analysis. 2026;114:104232. doi:10.1016/j.media.2026.104232

BibTeX
@article{Ouyang2026,
  author = {Ouyang, Jiarui and Wang, Yihui and Gao, Yihang and Xu, Yingxue and Yang, Shu and Chen, Hao},
  title = {GenAR: Next-scale autoregressive generation for spatial gene expression prediction},
  journal = {Medical Image Analysis},
  year = {2026},
  volume = {114},
  pages = {104232},
  doi = {10.1016/j.media.2026.104232}
}
Kostenlos starten →