Hacker News Reader

Comment by jxramos | Hacker News Reader

I don't get it, how is it that the simple question of "compared to what?" is just completely omitted and seemingly never even considered by so many authors? Do people just operate with plain dead end facts without context or curiosity about how normal is even characterized or distributed? Is the idea of a nominal case/measure/etc so foreign? I wonder what series of basic questions could be memorized that could be a tool for authors to ask themselves to elicit getting this information or at least stating that the answers are unknown.

View on HN

I’m confused…did you read the Science paper itself? In their _first paragraph_ they reference baseline genetic variation, and age-related mutation versus the mutations seen in cancer survivors and—as it turns out—astronauts, in the second paragraph. Lots more details and caveats on the limitations of the data elsewhere in the paper.

You seem to be reacting to this article, which is a popular summary of the findings. I don’t even think it’s a misleading article—it conveys the qualitative conclusion (elevated rates of nonstandard mutations), and answers to the questions you posed are available in the full scientific study…

The article links to a Nature paper. Are you reading a different paper?

https://www.nature.com/articles/s42003-022-03777-z.pdf

This is the closest thing I can find to a reference to a control in the first two paragraphs. The figures also suggest they did not have a control, so the use of the phrase “elevated rates” is baseless:

> Considering baseline genetic and extrinsic variability, the development of tools that permit the assessment of individual genetic susceptibility would improve risk stratification and long-term clinical management.

If funded, that still wouldn’t be a reasonable control.

Maybe I’m missing a key sentence or something, but I’ve read the paper, and it seems to boil down to:

“Mutations are bad, and this expensive piece of equipment measures mutations, so we stuck expensive astronaut blood in it, and the gauge pointed to a non-zero number”.

Later:

> Variants (SNP/ InDel) generated with this method were compared with a normal dataset using Archer’s analysis pipeline to distinguish noise from a true call. The normal dataset was created with sequencing data from seven young, healthy individuals.

From the article:

> in this astronaut cohort (median age 44 years, range 37–67),

I'd like to see also a comparison with 7 normal persons with ~45 years old. How many mutations is the expected value for a ~45 years old person?

Archer's analysis pipeline isn't a well defined statistical test. It's a UI that browses results from the type of machine they used:

https://www.enzymatics.com/news/archer-analysis-pipeline-upd...

They filtered the results with various p-values < 0.01. Their supplementary data:

https://www.nature.com/articles/s42003-022-03777-z#Sec9

doesn't say how many samples the machine produced in the first place, but it does say the filter resulted in 35 hits. That tells us absolutely nothing about the rate of mutations in the astronauts vs the general population, or even if astronauts produce more novel mutations on average vs. the general population.

On top of that, they seem to have the data to answer these questions, but don't report it anywhere:

Do zero-mission astronauts also show a "true call" in Archer's analysis pipeline?

They ran multiple samples from the same astronauts (before and after missions). On average, does the same astronaut have more/more-novel mutations after each mission?

These would require additional experiments or data, but seem obvious to check:

Do they have unusually high or low numbers of mutations vs. the general population at the same age?

If so, grouping the general population sample, what other professions also show a "true call" according to the pipeline?

Would have been nice to see a comparison to alternates who were otherwise qualified but did not fly shuttles.

Seems weird, but if they had actually done that (much better) experiment (which has a moderate to high chance of detecting no meaningful difference) it's quite likely the result would not be "interesting" enough to be published in a major journal.

With identical twins, one of each pair gets to fly, the other stays on the ground. They could also look for time dilation effects ;)

I'm not following several different aspects of your comment here. First of all, the paper is in Nature, not Science. Second, the first and second paragraphs do contain some relevant words, but nothing quantitative - literally the only quantitative number (not nomenclature number) in the entire first 2 paragraphs is CHIP being defined as >2% of something. Furthermore the reference to baseline genetic variation you mention is verbatim as follows: "Considering baseline genetic and extrinsic variability, the development of tools that permit the assessment of individual genetic susceptibility would improve risk stratification and long-term clinical management."

And I would argue that the article (not paper) does _not_ convey the qualitative conclusion of elevated rates of nonstandard mutations, because it doesn't reference the baseline at all! It doesn't even present it as a comparison, like "We found MORE mutations in astronauts than would be expected of non-astronauts" - it literally just phrases it as "We found mutations in astronauts", which, considering that non-astronauts are exposed to radiation and get cancer + epigenetic mutations as well, doesn't differentiate it from the normal expected observation.

You’re eliding an entire body of research with “>2% of something.” That assertion is cited, and links to a full study.

This is an observational study, and you may be analyzing it as a designed one. There was no randomization, no large scale matched control group per se—that wasn’t the intention. The intention is to analyze a specific population, and compare it to the existing literature on prevalence of mutations. That’s how these things are done, you have to use the citations given to understand the broader context of a study. One study is not useful on its own.

Also, there’s this—they did hypothesis tests against a control group:

How am I eliding anything? My statement is that the only quantitative number in the first 2 paragraphs is the ">2%" figure.

I'm also not criticizing the study, I'm criticizing the article and your defense of it. If "you have to use the citations given to understand the broader context of the study" - then if the point of a pop sci article is to effectively communicate the salience of the paper to lay audience, then that broader context needs to be communicated too.

No reasonable person reads “found genetic mutations in every astronaut” to plausibly mean “found bog standard genetic mutations everyone’s got.” This is pop science communication—journalism—not the actual scientific study. The goal isn’t scientific precision, it’s democratizing technical work

Really? I think everybody in the entire world knows somebody who's died from cancer, and the more educated of those people know that cancer is caused by genetic mutations. Also the theory of evolution is based on random genetic mutations - a very popular theory. I think a reasonable person would therefore know that everybody gets random genetic mutations over time.

To put it concretely, if the chance of a random gene mutation is 10^-4 to 10^-6 per gene per generation, and there are 37 trillion cells in a human body, each cell containing about 20k genes, (all numbers from quick google searches) then the odds of having NO genetic mutations occur in your body in one day would be (forgive my back of the envelope math, assuming a generation is 20 years): (1 - 10^-6)**(37 trillion * 20000 / 365 / 20). My calculator can't keep enough precision to make this nonzero. i.e., everyone has genetic mutations constantly which makes the title of the article completely uninformative.

There are some tricks to reduce the number of important mutations.

Germinal cells reproduce very slowly, so the eggs and sperm have fewer mutations.

Also, cells inside the guts have some cascade method. The cells near the wall reproduce very seldom, but the intermediate cells reproduce faster, and the inner cells reproduce even faster. The inner cells that are reproducing fast die or get washed away, so the big number of possible mutations is removed. And the slow reproducing cells near the wall create more cells to replace the intermediate cells that replace the inner cells.

Cool, that makes sense but I'd never seen it framed this way. (I knew about the slow rate for germ cells, but not gut.).

The article itself is full of caveats and concludes that their finding is inconclusive... Just some examples:

>Overall, further longitudinal studies are required to characterize CH and somatic mutational profiles in the context of space flight-associated stressors and their associated clinical impact. To date, there is no evidence of relevant CVD, cancer, or neurodegenerative diagnoses associated with this given astronaut cohort (current median age 62.5 years (IQR 60–67)). The lack of longitudinal samples from these same astronauts limits the assessment of clone stability, pathogenic potential, and prognostic value

But that the value of their report is in demonstrating the possibility of using stored archival blood samples in future studies:

>Thus, this study serves to address the feasibility of using bio-banked astronaut samples and demonstrate the importance of collaborations between NASA’s Human Research Program, Translational Research Institute for Space Health, Space Biology Program, NASA’s clinical support teams and corre- sponding data and biorepository branches,

Unfortunately, though, people far and wide are inevitably going to use this paper as evidence that space travel causes leukemia...

> The article itself is full of caveats and concludes that their finding is inconclusive... Just some example

It's a bad example. Mutations are known to cause various health problems like cancer. It's like randomly writing to the RAM, live: maybe you won't crash the computer immediately, but keep trying and you'll do damages.

The fact there's "no evidence of relevant CVD, cancer, or neurodegenerative diagnoses associated with this given astronaut cohort" shouldn't be surprising, given that astronauts are finely selected for perfect health among a large pool of candidate.

It's like saying "it's totally ok to do drugs or performance enhancer when you're a sports professional player" - no, they will eventually degrade the pro player health too!

Compared to a random person, it might just show later, as they have more of a health capital.

> Unfortunately, though, people far and wide are inevitably going to use this paper as evidence that space travel causes leukemia...

Given everything we know, yes it should, among other things - just like writing to RAM should eventually crash a computer

Mutations acquired during a lifetime are far more common that commonly believed; it's a key aspect of ageing. The astronauts in this study had an average age of 44, by which time they would be expected to have acquired mutations under normal, earthbound, conditions.

To address this, the study compares the number of mutations found in the astronauts with the number found in a broader population with mean age of 58 yrs. Notably, they do not conclude that more mutations were found in the astronauts.

If I were on a tenure review board, I would argue against treating Nature papers as peer reviewed publications. I keep seeing obvious “strong reject” clickbait from them, specifically.

This paper is a prime example.

Are you referring to the paper, or the summary of the paper which was linked to here?

The actual paper. It’s the top link in the body of the article.

Reply on news.ycombinator.com