Serious errors plague DNA tool that's a workhorse of biology | Hacker News Reader

Serious errors plague DNA tool that's a workhorse of biology

nature.com

98 pointsbookofjoe12 comments

I've been learning a lot about molecular biology recently, and am about to run some procedures related to plasmid isolation, gene cloning, and expression. Of note, the resources I've read, from AddGene, and The Bumbling Biochemist (excellent Youtube channel) emphasize sending your plasmids in for sequencing once complete, due to all the errors that can occur.

I am wondering if the plasmids the researchers here tested simply weren't sequenced. This wouldn't explain all classes of errors (like toxic protein production), but it may explain some. When you order from AddGene, they provide both the depositor's sequence results, if available, and their own results. The paper doesn't mention AddGene specifically; I'm curious how their plasmids compare to the ones tested.

Regarding toxic proteins: It seems like that would be a straightforward software addition to pass sequencing results through.

I figure the vast majority of plasmids are used for mundane stuff like GFP expression. Not really worth sending in for an analysis because everyone just redoes the experiment if it fails to express anyway.

Exactly! Everyone worth their salt sequences before, after and during insertion and expression.

I read this article expecting something new but it's just reporting that some people are lazy and do bad science.

This article is nonsense.

Before these companies like plasmidsaurus that do whole-plasmid sequencing for relatively cheap with nanopore, people generally only sequenced a region of interest using sanger sequencing. The rest of the plasmid was assumed to be mostly correct, as long as it grows on a bacterial resistance. As noted in the article, the rise of nanopore-based whole-plasmid sequencing has reduced a lot of these types of errors.

OK, this looks great. $15-60 for plasmid sampling!

This is good important work - similar to the stuff about antibodies that's been looked into. Hopefully as lab automation/analysis costs keep coming down it'll become more and more feasible to implement validation/characterization requirements into publication processes and increase reproducibility, and it'll be important to know where we're falling short currently.

Honestly, a crazy amount of biology and bioinfo works like that. Do any in silico analysis on NCBI accession genomes. I would say wast majority of them is either shoddy at best and at worst completely mislabeled.

Bioinfo tools? You scrape the surface of the reported results and see what is inside and a lot of stuff is either broken or work on a trust me bro principle.

Sequencing? In one technology results can be a bit different depending on the graphic card used and reported error rate is kinda shady.

Don't get me wrong people working with these stuff are amazing but the constant push for publishing more and more and the profit drive of the companies does a lot of bad stuff.

You are onto something there. I believe that there is a lot of good science being made. But there is a world of difference between even a well-conducted research paper and a 99.999999% SLA: we don’t have anything like that in biotech. Some will say that you just can’t. I say that we are not trying hard enough.

Back in the day I did some PhD research in horizontal gene transfer with plasmids. Really fascinating stuff.

Evolution used "it's not a bug, it's a feature" and it's super effective.

I wonder to what degree this affects DNA forensics tools.

Probably not at all. My understanding is that forensics uses either fragment analysis of pcr products, which depends on synthetic oligos, not plasmids. Those tests might be moving to NGS or long read sequencing, which also would not depend on plasmids. The reliability of forensic tests, good or bad, is another topic entirely.

Thanks!

Reply on news.ycombinator.com