RESEARCH NOTE 001 / DATASET & BASELINE
Building a genome–phenome atlas
The first step toward understanding living form is to connect a sequence record to a careful description of the organism it represents.
We assembled an initial atlas of public animal reference genome records and source-backed phenotype assertions. It makes a practical distinction between a genome file, the organism it came from, and what a source says about its species.
What exists today
| Resource | Initial coverage |
|---|---|
| Assembly records indexed | 14,329 |
| Species with phenotype profiles | 3,014 |
| Sourced phenotype assertions | 87,265 |
| Complete genome FASTA files downloaded and verified | 1 |
The profiles draw on AVONET, the Amniote life-history database, PanTHERIA, selected Phenoscape anatomy records, and a USGS species profile. Coverage is concentrated in birds, mammals, and reptiles. Different sources describe different properties; a field missing from a profile is unknown, not absent.
These are source-backed compilations, not 87,265 independently verified measurements. Some records are inferred or extrapolated; those are flagged. Species descriptions may combine sexes, life stages, environments, and studies. They usually do not describe the individual whose genome was sequenced.
One concrete genome–organism pair
We downloaded and checked the GCA_051020865.1 assembly for the spotted green pufferfish, Tetraodon nigroviridis. It contains 21 nuclear chromosomes totaling 342,830,278 bases, plus a 16,450-base mitochondrial sequence. This is one haplotype; it is not both copies of the animal’s genome.
The USGS species account, under the name Dichotomyctere nigroviridis, describes dorsal fin rays, anal fin rays, body form, markings, and habitat. We preserved the taxonomic connection and source attribution. These anatomical descriptions are species context, not measurements made on the sequenced specimen.
Can a first genomic representation predict traits?
An initial exploratory test matched up to 47 mammals to 3,604 published amino-acid gene-tree profiles. We evaluated body mass, longevity, gestation, and litter size while withholding entire taxonomic families. These historical comparative sequence inputs are separate from the current genome records in the atlas.
The gene-based model improved on a training-set mean, but the combined models did not reliably improve on both relatedness comparisons. For body mass, gene profiles reduced family-balanced squared error by 20.0%; a nearest-relative comparison reduced it by 47.2%. Those numbers describe prediction error in a small benchmark. They are not percentages of DNA decoded.
Finding: the atlas is a usable research starting point. This first prediction experiment does not establish a new biological mechanism or an advantage beyond ancestry.
Our follow-up anatomy test adds another strong competing explanation: ordinary body-size scaling. The result reinforces why a successful benchmark must do more than recognize related species.
Limits that matter
The atlas indexes assembly records rather than storing all public genomes. A single species may have several assemblies. Profile coverage is uneven, anatomy annotations can conflict, and source assertions have not all received record-level DNALYZER review. Comparative gene trees include sequences from evaluation species, so the pilot is not a fully independent test of learning from a previously unseen raw genome.
Next work should improve measurement context and test a defined anatomical contrast across independent evolutionary changes. A larger catalogue is valuable only if the added information helps discriminate between competing biological explanations.
Sources & research status
Data coverage and model results above are DNALYZER’s exploratory analysis, executed in September 2026. The inputs are public scientific resources; their authors did not endorse or conduct this analysis.
- NCBI Datasets — genome assembly records.
- Tobias et al., AVONET (2022) and dataset — avian morphology and ecology.
- Myhrvold et al., Amniote database (2015) — life-history traits.
- Jones et al., PanTHERIA (2009) — mammalian traits.
- Phenoscape Knowledgebase — source-linked comparative anatomy.
- USGS spotted green pufferfish account — species description.
- RERconverge public repository — example gene-tree inputs; this is not an official RERconverge analysis.
Publication policy: these notes share aggregate findings, source attribution, and limitations. Internal implementation and future candidate work are not included. Public summaries alone are insufficient to reproduce every analysis.
Explore the founding mission