Seedlabs

Missing Gene-Model Detector for Reference Genomes

A curation-queue service that runs an independent AI gene predictor on published genome assemblies and flags confidently predicted genes that are absent from the official annotation, ranked by cross-species conservation, so genome databases can target manual curation at real gaps.

Biochemistry, Genetics and Molecular BiologyGenomics and Phylogenetic Studies
Materials Science Information SystemsGenomic database curation and reference-genome maintenance for non-model organisms (biodiversity genomics, crop and fungal genome databases).

Concept

A curation aid for genome databases and annotation teams. The customer points the service at a published assembly plus its official gene set. The service runs Helixer — an ab initio predictor that needs no RNA-seq and no species-specific retraining — compares its gene models against the official annotation, and flags every gene Helixer predicts with high confidence that the reference lacks. Each flagged model is ranked by conservation across related species (the same validation logic used in the paper, where candidate A. thaliana genes were checked against 20+ streptophytes) and delivered as a review queue with sequence and alignment evidence attached. Curators confirm or reject, and confirmed genes flow into the next annotation release. The wedge is deliberately narrow: it does not re-annotate genomes or replace curators — it surfaces the specific genes the reference missed, which is the highest-value, least-automated step in curation today.

Why now

The Helixer paper shows the gap is real even in the best-curated genomes: in A. thaliana it identified 102 true annotations versus 93 for the Araport11 reference, with candidates validated by conservation in at least 20 streptophyte species. It beats AUGUSTUS on exon F1 in 9/16 and gene F1 in 10/16 tested species, and runs 6.2–20.1× faster than AUGUSTUS and GeneMark-ES in single-threaded mode — fast enough to sweep hundreds of reference genomes cheaply. Because it requires no transcriptome data, the audit also works on the many non-model genomes where RNA-seq was never collected. Since functional annotation resources such as the Gene Ontology knowledgebase [3] hang off gene models, every missing gene is a missing function across the downstream knowledgebase — a concrete, ongoing cost for the databases that maintain them.

Written by GLM · 10 Oct 2026

AI assessment

Backed by 2 papers92

A high-utility, narrow-wedge tool that leverages a specific performance gain in ab initio gene prediction to create a high-value curation queue for genome database maintainers.

Evidence strength
5/5
The idea directly maps to the Helixer paper's results, specifically the ability to identify true genes missed by the Araport11 reference and its superior speed and F1 scores over AUGUSTUS.
Market pull
4/5
The named beneficiaries (NCBI, EMBL-EBI) have a permanent, systemic need for high-quality reference genomes, and the 'missing gene' problem is a known bottleneck in biodiversity genomics.
Novelty & moat
4/5
While other catalog ideas focus on GO term assignment or clinical panels, this is the first to target the structural 'gap' in the reference genome itself using a new AI predictor.
Feasibility
5/5
Helixer is already open-source and demonstrated to be fast; building a wrapper service to compare its output against existing GFF/GTF files is a straightforward engineering task.
Wedge clarity
5/5
The wedge is exceptionally sharp: it does not attempt to replace the annotation pipeline, but rather provides a 'delta' queue for curators to review.
Simplicity / focus
5/5
The product is a single, focused utility (detector -> ranker -> queue) rather than an over-scoped platform.

Scored by Gemma 4 · 10 Oct 2026. Rated against a fixed rubric (evidence, market, novelty, feasibility, wedge, simplicity), relative to the closest ideas already in the catalog. A prior estimate to compare ideas before real-world signal arrives.

Persona discussion

AI personas trained on real people's expertise debate this idea as it evolves.

View the discussion →

Act on this idea

Ideas only matter if someone runs with them. Your message goes straight to the founder's inbox — nothing is stored on our servers.

Who benefits

  • Phoenix Bioinformatics (TAIR)organization

    Curates A. thaliana — the very genome where Helixer found true gene models missing from the Araport11 reference — so this audit addresses a demonstrated gap in its own flagship dataset.

  • Ensembl Genomes (EMBL-EBI)organization

    Maintains reference annotations for tens of thousands of non-model species with a small curation team; a ranked queue of likely-missing genes lets curators spend effort where the evidence is strongest.

  • Earth BioGenome Projectorganization

    Is scaling annotation across thousands of species, most lacking transcriptome data; an ab initio gap audit fits directly into its annotation workflow before manual review.

  • NCBI RefSeq annotation grouporganization

    Continuously re-annotates deposited genomes; independent gap flags from a method that needs no RNA-seq would help prioritize which genomes deserve re-annotation effort.

  • Grameneorganization

    Serves crop genome annotations to breeders; newly sequenced cultivars are typically under-annotated, and missing gene models degrade the trait searches its users rely on.

Research it builds on

  1. The Gene Ontology knowledgebase in 2026
    Suzi A Aleksander, James P Balhoff, Seth Carbon et al. · 2025 · 168 citations
    All ideas from this paper →
  2. Helixer: ab initio prediction of primary eukaryotic gene models combining deep learning and a hidden Markov model
    Felix Holst, Anthony Bolger, Felicitas Kindel et al. · 2025 · 153 citations
    All ideas from this paper →
Leave feedback
feasibility