[fjolubla]

PR Checklist for a new package submission

• The package does not exist already in the community archive, also not with a different name.
• The package title in the POSEIDON.yml conforms to the general title structure suggested here: <Year>_<Last name of first author>_<Region, time period or special feature of the paper>, e.g. 2021_Zegarac_SoutheasternEurope, 2021_SeguinOrlando_BellBeaker or 2021_Kivisild_MedievalEstonia.
• The package is stored in a directory that is named like the package title.

---

• Samples that already have been published previously, and got re-analysed (e.g. re-sequenced) for the now packaged publication, have a modified Poseidon_ID of the form <Original Poseidon_ID>_<Initials of the main author>_<Year>. Re-analysed versions of I1685 (Lazaridis et al. 2016) should, for example, be assigned the IDs I1685_IL22 (Lazaridis et al. 2022) and I1685_IL25 (Lazaridis et al. 2025).

---

• The package is complete and features the following elements:
  • Genotype data in binary PLINK format (not EIGENSTRAT format).
  • Genotype has been provided by the original authors of the publication describing the data.
  • A POSEIDON.yml file with not just the file-referencing fields, but also the following meta-information fields present and filled: poseidonVersion, title, description, contributor, packageVersion, lastModified (see here for their definition)
  • A reasonably filled .janno file (for a list of available fields look here and here for more detailed documentation about them).
  • A .bib file with the necessary literature references for each sample in the .janno file.
• Every file in the submission is correctly referenced in the POSEIDON.yml file and there are no additional, supplementary files in the submission that are not documented there.
• Genotype data, .janno and .bib file are all named after the package title and only differ in the file extension.
• The package version in the POSEIDON.yml file is 1.0.0.
• The poseidonVersion of the package in the POSEIDON.yml file is set to the latest version of the Poseidon schema.
• The POSEIDON.yml file contains the corresponding checksums for the fields genoFile, snpFile, indFile, jannoFile and bibFile.
• There is either no CHANGELOG file or one with a single entry for version 1.0.0.

---

• The Publication column in the .janno file is filled and the respective .bib file has complete entries for the listed mentioned keys.
• The .janno file does not include any empty columns or columns only filled with n/a.
• The order of columns in the .janno file adheres to the standard order as defined in the Poseidon schema here.
• The .janno and the .ssf files are not fully quoted, so they only use single- or double quotes ("...", '...') to enclose text fields where it is strictly necessary (i.e. their entry includes a TAB).

---

• The package passes a validation with trident validate --fullGeno.

---

• Large genotype data files are properly tracked with Git LFS and not directly pushed to the repository. For an instruction on how to set up Git LFS please look here. If you accidentally pushed the files the wrong way you can fix it with git lfs migrate import --no-rewrite path/to/file.bed (see here).

[nevrome]

Excellent job, @fjolubla! I only saw some minor details. Maybe you could quickly address these and then I think we can merge this.

[nevrome]

What does the uncertainty level not independently verified mean?

[nevrome]

I also think the Date range from Feldman et al. (2026) is a unnecessary.

[fjolubla]

I took the date ranges from supplementary Table S1A but could not verify whether they are reporting 68.2% (1σ) or 95.4% (2σ) confidence intervals, since the janno documentation prefers 2σ. But maybe it would be better to report this in the readme instead?

[nevrome]

I looked into the supplementary material now and understand your confusion. There are three columns that mirror the information we would like to have in the Poseidon package (in years BP):

Date_mean (BP; based on calibrated, corrected C14 2-sigma when available)
Date_LB (BP; based on calibrated, corrected C14 when available)
Date_UB (BP; based on calibrated, corrected C14 when available)

It's not entirely clear if "corrected C14 2-sigma when available" means that it's always the 2-sigma range, or only for the dates the authors calibrated themselves (Table S1B). I tried to look at the source publications they list, but ran into issues there as well:

I could not find WK-25215 and WK-25223 in Fenner et al. 2015, because that only reports WK-25217, WK-25218, WK-25224, WK-25226, and WK-25227. They mean 10.1016/j.jasrep.2014.11.008, right? They only cite Fenner et al. 2021, so I had to guess. I must be missing something here.

I then tried Stantis et al. 2016 in search for i.A. Wk-17451, but they report different probabilities, sometimes slightly below 95.4% for some individual dates. Maybe they only consider the biggest high density region?

I think generally it's customary to give the 95.4% (2σ) confidence interval if nothing else is stated. Fenner et al. 2015 do this for the dates that I can see and Stantis et al. 2016 seem to do it at least for the samples with no multimodal post-calibration probability distribution. I think we can reduce Date range from Feldman et al. (2026), Table S1A; uncertainty level not independently verified; to Date range as reported by Feldman et al. (2026), Table S1A;. This is sufficient.

[stschiff]

Thanks to both for the submission and the review. Great addition!

[fjolubla]

I was still waiting to hear back about some potentially missing genotype files, which is why I hadn't updated/addressed comments it until now. Unfortunately, i still havent got any update about those files, so if they become available, I may need to update the package again, the affected samples are excluded in this version. But in the meantime i have addressed your comments @nevrome

[nevrome]

The .ssf file was missing some of the information we usually require, so I added that. I will merge this now 👍