henhaobio
Sign in

bioRxivDevelopmental BiologyPreprint

gene2fish: a multi-gene spatial expression browser for in situ hybridization data across zebrafish development

Vera JansseniD, Wellington D. F. Rutes, Leandro LimaiD, Kyle AwayaniD, Benjamin Giovanni IovinoiD, Rafael Gómez-SjöbergiD, Löıc A. RoyeriD

PDFDOICodeData
In-depth full-text digest available · Jump to digest ↓

Abstract

Single-cell RNA-sequencing (scRNA-seq) atlases of zebrafish development have catalogued marker genes for hundreds of cell types, but transcriptomic data alone cannot reveal their expression patterns in the embryo. The systematically annotated Thisse high-throughput whole-mount in situ hybridization (ISH) collection at ZFIN provides spatial expression data across twelve developmental stages, yet the ZFIN web interface is organized around individual genes, making side-by-side cross-gene comparison laborious. We present gene2fish, a web application that displays ISH images from the Thisse dataset, linked directly to their ZFIN records, in a gene-by-developmental-stage grid. With gene2fish, researchers can quickly explore and validate the spatial expression of marker gene lists, and find the genes annotated to specific anatomical structures across developmental time. The application is part of Zebrahub and is freely available at https://zebrahub.sf.czbiohub.org/gene2fish.

In-depth summary

Claude Opus 5.5 · xhigh effort · 8 Nov 2026 · from the full text of version 2 (21 Oct 2026) · CC BY 4.0

gene2fish arranges 53,595 Thisse in situ hybridization images of 8,263 zebrafish genes into a gene-by-stage grid, with anatomy-term search for validating scRNA-seq markers.

Zebrafish scRNA-seq atlases yield marker lists without spatial context, spatial transcriptomics remains specialized and scarce at later stages, and ZFIN presents the largest systematic ISH resource (Thisse collection) only gene by gene. gene2fish aligns many genes on a shared developmental-stage axis and lists genes annotated to chosen anatomical structures, easing marker validation for non-specialists.

Key results

  1. ShownBuilds a per-image index of the Thisse ZFIN in situ hybridization (ISH) collection by joining ZFIN public reports. It covers thousands of zebrafish genes, and almost all images are whole-mount mRNA ISH of wild-type AB/TU animals.

    53,595 images of 8,263 genes from five ZFIN direct data submissions (snapshot downloaded April 14, 2026). Whole-mount preparations make up 99.6%; the rest comprise 166 sectioned samples, 14 imaged alive and 30 with no recorded preparation type. 3,850 genes have at least one stage-resolved image. · gene2fish application; Data source

  2. ShownCoverage is sparse per gene and uneven across development. Most genes have a single image, coverage is concentrated in somitogenesis and pharyngula stages, and a large set of images, mostly of pooled mixed-stage embryos, has no assigned stage.

    Images per gene: median 1, mean 6.5, maximum 144; twelve stages from sphere to day 5 shown; highest coverage from 1–4 somites through High-pec; germ-ring, shield and pec-fin each have only one or two genes; 4,909 images placed in an unspecified-stage row. · gene2fish application; Developmental stages; Figure legends

  3. ShownAnatomy search over ZFA ontology terms returns the genes annotated to a chosen structure. By default it includes substructures, it intersects results when several terms are selected, and it never counts 'expression not found' annotations.

    227 annotated terms selectable; substructure expansion uses 5,389 is a/part of relations; median 45 genes per term (mean 149.2, maximum 3,181); for example, 'somite' returns 777 genes and 'diencephalon' returns 647 genes. · Anatomy annotations and search; Figure legends

  4. ShownQueried genes form the columns and developmental stages the rows, with ranked representative images, a stage-window filter, shareable URL-encoded views, PNG export and a full-resolution viewer showing ZFIN and human ortholog metadata.

    Images ranked by preparation, then positive annotation count, then ZFIN image ID; 1, 3, 6 or 10 images per cell; probe quality rated 0 to 4 but missing for 30% of images; viewer lists ZFIN gene ID, stage, ZFA terms, probe, fish line, human ortholog, orthology-inferred diseases and UniProt ID. · gene2fish application; Image selection and display; Figure legends

  5. ClaimedThe authors argue that the grid lets researchers check scRNA-seq cluster markers and compare expression domains at matched stages at a glance, in place of gene-by-gene ZFIN lookups, and that it opens the data to researchers outside the zebrafish field.

    Illustrative scenario: surveying 10–20 cluster markers in one browser tab instead of 10–20 sequential ZFIN queries. Shown only through example queries (cxcr4a, prickle2b, shhb, pax2a) and a 78 s walkthrough video; no user study or timing benchmark is reported. · gene2fish application; Discussion; Figure legends

Methods, data and reproducibility

A Python script joins ZFIN public data reports (April 14, 2026 download) into one record per image. A FastAPI backend serves gene, stage, anatomy and image queries from this in-memory index to a React/Vite frontend, which renders a gene-by-stage ISH image grid with ZFA ontology-aware anatomy search.

  • Thisse et al. high-throughput ISH collection at ZFIN: images and expression annotations from five ZFIN direct data submissions (mRNA ISH on wild-type AB/TU embryos and larvae) · zebrafish (Danio rerio) · n = 53,595 images of 8,263 genes · ZDBPUB-010810-1; ZDB-PUB-040907-1; ZDB-PUB-051025-1; ZDB-PUB-080220-1; ZDB-PUB-080227-22
  • Archived per-image metadata index built from ZFIN reports downloaded April 14, 2026, with anatomy hierarchy, Supplementary Table 1 and figure scripts · zebrafish (Danio rerio) · n = 53,595 images · doi:10.5281/zenodo.23378512
  • ZFA anatomy ontology is a/part of relations used for substructure-aware anatomy search (ZFIN release of September 28, 2026) · zebrafish (Danio rerio) · n = 5,389 relations; 227 annotated terms
  • ZFIN human orthology, orthology-inferred disease, UniProt and gene alias reports linked to each image record · zebrafish genes with human orthologs
  • gene2fish v0.14.2 source code archive · doi:10.5281/zenodo.23380713
Methods & tools
Thisse high-throughput whole-mount mRNA in situ hybridization, ZFIN public data reports, Zebrafish Anatomy Ontology (ZFA), zfin image metadata extractor.py (index-building Python script), FastAPI (Python 3.11), React 18, Vite, pytest, Playwright, matplotlib, UniProt and Disease Ontology cross-references, Claude (Anthropic), assisting manuscript editing and software development
Software & environment
Python 3.11 FastAPI backend serving an in-memory index through a REST API; React 18 single-page frontend built with Vite; both packaged in one container image, with the index built at image build time; release v0.14.2 under MIT license.
Reproducibility gaps
The live application rebuilds its index from the current ZFIN reports on each redeployment, so its counts may differ from the archived April 14, 2026 snapshot. Images come from a Biohub bucket, with ZFIN as the fallback source.
Limitations (6)
  • Coverage is restricted to genes assayed in the Thisse high-throughput collection. Genes outside it appear as empty grid cells even when other ISH datasets for them exist at ZFIN. · stated by the authors
  • Stage coverage within the collection is uneven (germ-ring, shield and pec-fin have one or two genes each), and 4,909 images, mostly of pooled mixed-stage embryos, can only be shown in an unspecified-stage row. · stated by the authors
  • Other spatial modalities, such as smFISH, RNAscope or spatial transcriptomics resources like the MERFISHEYES atlas, are not yet integrated. · stated by the authors
  • The tool only displays images and annotations. It does not classify expression patterns automatically, so interpretation is left to the user. · stated by the authors
  • The anatomy relations cite a ZFIN release dated September 28, 2026, but the metadata were downloaded April 14, 2026. The ontology hierarchy and the annotations in the archived snapshot may therefore come from different versions. · noted in this summary
  • The claimed workflow and accessibility benefits for non-specialists are not tested with user studies or with comparisons against ZFIN's own expression search. · noted in this summary

Read this if you validate zebrafish scRNA-seq marker genes against in situ expression patterns or need the genes annotated to an anatomical structure across developmental stages.

Citations
0

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.

Related papers

Similarity Map
SelectedSame fieldCross-field
THISFarnsworth '19Liu '22Aponte-Sant '26White '17Guo '23Wan '26
67% semantic matchDevelopmental Biology
Dylan R. Farnsworth, Lauren M. Saunders, Adam C. Miller·Developmental Biology·Nov 27, 2019· 276 citations