Deep learning to predict the lab-of-origin of engineered DNA

Well, consider that if airplanes were 99% safe, nobody would fly. And the way I understand it (not very well), the lab-of-origin discovery is based on design choices rather than inherent factors. This means that it could be copied to create false positives? Hopefully someone more educated on the subject can elaborate.

Neither - they are talking plasmids, which are a precursor to both. Plasmids are circular DNA that are used in cloning. Cloning is the step-by-step process by which a construct is engineered (e.g. promoter and gene sequence placed in the correct orientation, along with a selectable maker in traditional GM). Plasmids are amplified in bacterial systems to generate loads more, and a very stable. Good for storing and passing around to other labs. Often the same plasmid backbone is used in multiple constructs, hence why the approach in the paper works. You could not use this approach to detect which lab CRISPRd an organism (the tell-tale signs are not retained), but it would have some effectiveness in traditional GM if the insert were sequenced, as some of the lab-specific cloning decisions and mutations could be present.

> I wish papers that get published contained a section with the reviewers' notes.

At least the journals published by the European Geosciences Union publish the whole review process in the open. Reviewers' comments, and authors' responses to them.

https://www.egu.eu/publications/open-access-journals/

(When you open the view to any article in any of those journals, you need to click on the "Peer review" tab, and you get to see the original manuscript as "Discussion paper", and all of the review process.)

Current Anthropology does this - see the 'comments' starting on page 407 of this paper, for example:

https://www.journals.uchicago.edu/doi/pdfplus/10.1086/698936

I have no idea if this is standard in anthropology. I am not an anthropologist, nor do i play one on television.