# Mammal chromosome counts: ProteinIQ's comparative analysis

Analysis date: September 19, 2026. This is our descriptive analysis of published data, not new experimental measurements or a peer-reviewed study. The question is how the familiar human, cat, horse, and dog chromosome counts compare with a reproducible sample of mammals.

## Findings

Our main sample contains 465 species labels. The median female diploid count is 44; the most frequent count is also 44 (62 species). Of these species, 290 (62.4%) have fewer than 46 chromosomes, 13 have exactly 46, and 162 have more. At the dog/gray wolf reference count of 78, 458 (98.5%) are below, two are equal, and five are above.

These percentages describe the included records, not all living mammals. No population-weighting, phylogenetic correction, uncertainty interval, or claim of biological novelty is intended.

## Inputs and attribution

1. **Animal Chromosome Count Database (ACC)**, Cristian Román-Palacios, Cesar A. Medina, Shing H. Zhan, and Michael S. Barker. March 24, 2021 release, version 1.0.1, archived March 25, 2021 at [Zenodo, DOI 10.5281/zenodo.4635457](https://doi.org/10.5281/zenodo.4635457). We extracted all 1,327 Mammalia records from `ACCDB_Mar24.csv`, obtained from a pinned GitHub commit. The original file's MD5 agrees with the archived file's checksum. Zenodo identifies its license as CC BY. `acc-mammals.csv` preserves source columns and adds `source_record`, the one-based data-record position in the original CSV, excluding its header. The original archive has 14,524 animal records; only mammals enter this analysis.
2. **Blackmon, Justison, Mayrose, and Goldberg (2019)**, [Meiotic drive shapes rates of karyotype evolution in mammals](https://doi.org/10.1111/evo.13682), *Evolution* 73:511–523, Table S1. We extracted species names and the numerical female haploid counts for 1,060 records from `supp.table.1.docx` in the Europe PMC supplement. The paper defines these counts to include autosomes and sex chromosomes. `blackmon-table-s1.csv` retains identifiers for the two XML tables and row positions, including their headers. The 1,059 species in the paper's phylogenetic analysis are a different denominator. The associated data archive is [Dryad, DOI 10.5061/dryad.rg5s170](https://doi.org/10.5061/dryad.rg5s170); we used the article supplement, not a downloaded Dryad file. The article's license is CC BY-NC 4.0. We extract numerical facts and names for analysis and do not redistribute the article, original table document, formatting, or other supplemental files.

Both sources were retrieved September 19, 2026. Exact URLs, source identifiers, and SHA-256 checksums are in [provenance.json](provenance.json). The input CSVs are extracts of their data; the audit, summaries, comparisons, and distribution are our derived outputs.

## Inclusion and calculation

We replace underscores in species names with spaces and trim surrounding whitespace. We make no synonym substitutions or modern taxonomic revisions. Duplicate normalized names fail validation. This produces 663 shared names, 664 ACC-only names, and 397 Blackmon-only names, or 1,724 distinct labels in the audit. Unmatched names can reflect taxonomy as well as coverage; the union is not a count of currently accepted species.

For the main sample, an ACC record must have exactly one distinct positive integer diploid count, equal to twice the Blackmon female haploid count. Repeated identical counts are deduplicated. If ACC also gives an explicit haploid count, it must likewise have one distinct value equal to Blackmon's. An absent ACC haploid count is allowed. We do not use ACC's rounded or inferred haploid fields, average multiple counts, infer missing zeros, or substitute manually reviewed values.

Each included species label contributes one observation, irrespective of how often it was studied. The four comparison species are Felis catus, Homo sapiens, Equus caballus, and Canis lupus. Dogs and gray wolves share the latter entry and are counted once. The analysis uses the female diploid complement for consistency with Blackmon's definition, not a claim that both sexes always have identical totals.

The mutually exclusive decisions, in order of evaluation, are:

| Decision | Records | Meaning |
| --- | ---: | --- |
| acc_only | 664 | No exact Blackmon name match |
| blackmon_only | 397 | No exact ACC name match |
| missing_acc_diploid | 1 | Shared name with no explicit ACC diploid count |
| count_disagreement | 71 | Twice Blackmon's count is absent from ACC's diploid list |
| multiple_acc_counts_including_reference | 121 | ACC lists multiple distinct counts, one agreeing with Blackmon |
| acc_haploid_conflict | 5 | Diploid agreement but conflicting explicit ACC haploid information |
| included | 465 | All main-sample criteria met |

A record can have additional problems beyond its first exclusion reason. For example, the multiple-count category is decided before the haploid consistency check. These categories do not distinguish all biological variation from data-entry errors.

For each reference count, the proportion below is `100 × number strictly below / sample size`. Ties are reported separately and are not included in the numerator. Medians use the middle observation, or the mean of the two middle observations for even sample sizes. Modes include all tied most-frequent values. Histogram bins are inclusive integer intervals of width ten, covering 10–109; the script fails if any main-sample value falls outside them. `distribution.csv` retains exact, unbinned counts.

## Sensitivity and limitations

| Treatment | Species | Median | Below human count of 46 |
| --- | ---: | ---: | ---: |
| Main criteria | 465 | 44 | 290/465 (62.4%) |
| Ignore explicit ACC haploid conflicts | 470 | 44 | 291/470 (61.9%) |
| Also allow multiple ACC counts if one agrees | 591 | 42 | 366/591 (61.9%) |
| Entire Blackmon table without the ACC screen | 1,060 | 42 | 630/1,060 (59.4%) |

Broader treatments use twice Blackmon's female haploid value. They are alternative selections, not repaired or independently validated datasets. Their median of 42–44 and human comparison of 59.4–62.4% show how results change under these specific choices; these ranges are not confidence intervals.

Agreement across compilations does not establish independent biological replication. The compilations can share underlying literature and therefore errors. We did not check the primary cytogenetic evidence for every included record. Selecting for agreement can preferentially omit variable species. It also changes the taxonomic composition; the main sample includes 101 bats and 93 rodents, for example. Source order labels are retained without taxonomic updates. Neither the sample nor its exclusions represent a random survey of mammals.

We separately reviewed illustrative anomalies and variable records, recorded in [manual-review.csv](manual-review.csv). This targeted review is not an estimate of either database's error rate. It does not alter the main sample. In particular, we did not turn single-digit ACC entries into presumed multiples of ten. Primary sources can explain a discrepancy, but a systematic repair would require a separate record-by-record review.

## Reproduction and files

Download this directory's files together. Python 3.10 or newer is sufficient; no third-party libraries are required.

```sh
python analyze.py
python analyze.py --check
```

The first command regenerates four outputs from the frozen CSV inputs; the second checks byte-for-byte reproducibility without writing. Input checksums are always verified. To retrieve and re-extract the recorded sources, with checksum checks before accepting them:

```sh
python analyze.py --refresh-inputs --check
```

The Europe PMC endpoint returns a ZIP containing `EVO-73-511-s001.zip`, which in turn contains Table S1. We verify the nested supplement and DOCX bytes, not the outer transport ZIP, whose packaging can change. A source checksum mismatch stops reproduction rather than silently accepting a new version.

- [analyze.py](analyze.py): extraction, inclusion rules, calculations, and output verification.
- [acc-mammals.csv](acc-mammals.csv) and [blackmon-table-s1.csv](blackmon-table-s1.csv): frozen numerical inputs.
- [species-audit.csv](species-audit.csv): all 1,724 labels with original values, source positions, references where available, and decisions. Filter `decision` to `included` for the 465 analysis records.
- [summary.json](summary.json): full statistics, bins, source-order composition, and sensitivity results.
- [distribution.csv](distribution.csv): exact chromosome counts and species frequencies.
- [comparisons.csv](comparisons.csv): counts below, tied, and above each familiar reference count.
- [manual-review.csv](manual-review.csv): targeted review notes and primary-source links.
- [provenance.json](provenance.json): source versions, identifiers, and hashes.

Suggested attribution: ProteinIQ (2026), *Mammal chromosome counts: comparative analysis of ACC v1.0.1 and Blackmon et al. (2019) Table S1*, analysed September 19, 2026. Cite the two original data sources alongside our analysis when reusing the findings.