LoginSign Up
← Back to all sparks
P

phyloatlas

INFRA · APIS
Velocity0.0

Access to the 'Phylo-Species Atlas' of Empirical Phylogenies

github.com/franciscorichter/phylo-species-atlas ↗

An atlas of the tree of life that keeps publishing what it got wrong, and stopped shipping the trees it does not own.

phylogeneticsresearch-datadata-provenanceopen-sciencereproducibilitylicensing
◆Current state
This is a curated deposit of species-level phylogenies — 264 trees across 62 partitions, 247 of them time-calibrated — paired with a species-name dictionary of roughly 638,000 standardized labels and per-tree provenance. It is moving through peer review at Methods in Ecology and Evolution, and the release stream is essentially a public erratum log: eight releases in five weeks, each reconciling the deposit against source papers, the manuscript, and its own metadata. The most recent replaced the turtle canonical with a 274-tip ultrametric chronogram after the previous tree failed time-calibration integrity checks.
◆Where it's heading
The project's defining decision is that it archives the recipe rather than the corpus. Since 1.0.3 the Zenodo deposit holds metadata, provenance, and standardization code while the tree files live at their original sources, and the same principle was applied again when the TimeTree-of-Life was removed at the TimeTree project's request and reduced to a citation. What makes the correction log unusual is its direction: partitions keep getting reclassified from dated to undated as verification confirms the source papers described chronograms they never deposited. The atlas is being built to be honest about archival uncertainty rather than to maximize coverage.
◆Prediction
The release cadence is driven by the manuscript review cycle, so expect corrections to continue until MEE acceptance and then slow sharply; the open thread most likely to produce the next one is the remaining recoverable archival uncertainty flagged across non-Condamine dated source trees.

◆Recent moves

  1. 1mo ago

    v1.0.8 — turtle chronogram canonical; MEE integrity pass

    The turtle canonical becomes a 274-tip ultrametric chronogram, replacing a 287-tip substitution tree that failed calibration integrity, with the change reconciled across metadata, provenance, estimates, and the supplementary table. The fungi count is also aligned to the shipped 1,602 species-level tips after 22 duplicate-species tips were identified in the raw tree. Both are the same species of fix the whole release run has been making: the shipped data and the claims about it converging.

    View source ↗
  2. 1mo ago

    v1.0.7 — restore species-name dictionary

    A correction to the previous release's own correction: the TimeTree removal had over-pruned the flat species-name dictionary, and 637,619 standardized labels are restored on the reasoning that a list of Linnaean names is not TimeTree's tree. The tree itself, its name shard, and membership rows stay out. It is a useful demarcation of what de-redistribution actually covers.

    View source ↗
  3. 1mo ago

    v1.0.6 — data-integrity corrections + TimeTree de-redistribution

    Eight partitions that had been shipping substitution or topology trees are replaced with the published time-calibrated chronograms, bryophytes are reclassified as undated because none was ever deposited, and the dated count settles at 247 of 264. The TimeTree removal here applies the redistribution principle set in 1.0.3 to a specific upstream request rather than establishing it.

    View source ↗
  4. 2mo ago

    v1.0.5 — consistency corrections

    Coverage numbers are corrected against the submitted manuscript, including a sensitivity bound moving from 6.6% to 5.4% and fish coverage restated as the shipped 11,638-tip molecular subset rather than the much larger imputed source tree. A double-counted crocodilians canonical is removed from the tree decomposition. Distinguishing what ships from what a source tree claims is the recurring discipline of this project.

    View source ↗
  5. 2mo ago

    v1.0.4 — data corrections + canonical succession (Supplementary Table S7)

    Four partitions are reclassified from dated to undated after verification against source papers found phylogenomic topologies with no chronogram deposited, in one case a chronogram described in a figure but never archived. Corrections that reduce headline coverage are the strongest evidence that the curation heuristics are being checked rather than trusted.

    View source ↗
  6. 2mo ago

    v1.0.3 — title alignment, LICENSE, three-category framework, sensitivity bounds

    ⚡ SPARK

    The release that decided what this artifact is. The Zenodo deposit stops holding standardized Newick files and starts holding the methodological recipe — metadata, per-tree provenance, fetching and standardization code — with the trees themselves left at the repo, the live atlas, and their original publication sources. Every redistribution decision since, including the TimeTree removal, follows from this.

    View source ↗