Hacker News Reader

> However I'm not in medical research, and even if I was it would have been impossible to get funding for researching the myriad of root causes

Having been a software engineer my whole career and then spending a few years in cancer research, I can tell you there is a big shortage/demand for senior software engineers and there are many funded PIs open to bringing on senior engineers to direct their own projects.

I'd encourage you to take a look as we need more good people!

Can you provide more detail? That sounds interesting.

Sure I can mention my experience. It's a common path I see on https://www.reddit.com/r/bioinformatics/ or https://bioinformatics.stackexchange.com/. For me I moved to a new state (wife's job) and checked out Craigslist to look for interesting jobs. The local university Cancer Center had openings for experienced software engineers. No bio background but they had great bio people and I could help them analyze their data, take their R and Python code and clean it up, make it faster, add tests, scale to the cloud, etc. I kept a bio notebook on the side and attended talks and lectures, spent time with people on the bench, read books/textbooks etc, and gradually picked up a lot of the bio (it's a massively big domain—OOM bigger than I realized). It's a great combo when you have people with bio experience + people with software engineering experience. I was able to help lots of teams with various technical challenges, start my own projects, and met a lot of teams throughout the country + world doing interesting research and am still involved with various research endeavors.

A key challenge in biology now is the huge amount of data and wrangling that requires smart software, so someone with strong software engineering skills can be the difference between publishing a paper in 3 months vs 3 years. If you search any of the many top tier bio research orgs (Allen, MD Anderson, Broad) in the USA you'll see lots of openings for software engineers without too much consideration for prior bio experience.

I often get irritated when people treat "correlation does not imply causation" as a slam-dunk refutation of a complex argument, although you do rescue yourself from this by shifting to the use of the word "proof". In the sense of "logically entails", or "proves", then it's trivially true that correlation does not imply causation. In the sense of "is evidence supporting"---which is nearly universally (outside of Math and Philosophy departments) the way that the word "implies" is used in English---correlation absolutely implies causation. It's a good heuristic for deciding where to focus research dollars, if only to uncover a more complex causal relationship. The absence of correlation is one of the most commonly used proofs in science. Surely its presence can be used as good evidence, if only because we have to operate in a universe where the excluded middle isn't self-evident.

Sure. Nobody (seriously) believes that correlations can’t be suggestive of a causal relationships....but they do need to be followed up.

It’s more that the AZ field has found this particular correlation over and over and over again. Despite millions (possibly closing in on billions) of dollars, however, it’s proven difficult to turn it into a causal link. A-Beta antibodies repeatedly fail, both before and after the plaques appear. Various types of inhibitors don’t work. If a method does something to amyloids, someone has probably tried it already, often in an expensive and ill-advised trial. This is only barely hyperbolic: https://blogs.sciencemag.org/pipeline/archives/2018/06/12/an...

Despite millions (possibly closing in on billions) of dollars...

Oh, we're definitely into billions. Per year, and not total over time.

This year alone, the NIH is spending $2.8 billion on Alzheimer's research. And they are not the only source of research money.

Almost all of that goes to research based on the amyloid hypothesis.

I thought there's no way that ~10% of the NIH budget would be thrown into that particular money pit, but lo and behold!

https://www.nia.nih.gov/sites/default/files/2019-07/FY21-NIA...

I actually was surprised to read the following sentence in there, While amyloid continues to be a target, 26 of the 41 pharmacological trials supported by NIA are focused on other targets.

That's still far too much being thrown into a theory with a history of failure. But there is a shift there. Perhaps being criticized in places like Nature and Scientific American had an effect? One can hope!

Your irritation notwithstanding, a complex argument that falls apart to "correlation does not imply causation" is a flawed argument. And humans so consistently err on the side of leaping to conclusions in advance of the facts, that we need to consciously counterbalance by reminding ourselves of the fact.

Yes, correlation is suggestive of causation. Yes, it suggests an often fruitful line of research. But while pursuing that line of research it is important to remain aware that it isn't proof. And to not shut off other lines of research.

In this light, it is worth reading http://calteches.library.caltech.edu/51/2/CargoCult.htm again. Science isn't supposed to be easy. Scientists have had to learn the hard way how to be honest with themselves. Some fields have done better on that than others.

Medicine's record on this front has generally been abysmal. And their failure to be honest with themselves about the limits of their own theories in this case has had an opportunity cost measured in trillions of dollars spent on millions of tragically preventable lingering deaths.

In the face of that, it is wrong to argue that their mistake was a defensible jump in reasoning that people shouldn't be blamed for. Instead we should notice that a logical error really can have a cost measured in trillions spent on millions of tragically preventable deaths. And figure out how to not do that again.

You're begging the question. My core point was that "correlation does not imply causation" doesn't "refute" anything but logical postulates---the problem with something that can be so dismissed isn't that it's logically flawed but that it isn't science. Which I think is closer to what Feynman was describing than what you're implying. When you say "science isn't supposed to be easy" (I think most people already know that, but thanks), you're supporting my position: saying "correlation is not causation" isn't helpful. In itself, it's either misleading or trivial. And it deserves a standard response: "Thanks, Professor. What's your experimental design to investigate that? Competing hypotheses? Evidence supporting your position?" Yes, science isn't supposed to be easy, and that applies to disputing theories and evidence, too. I have my own problems with the release in the OP, but are you seriously claiming that it's based on Cargo Cult Science? You know, or should know, that there's more than mere correlation in the evidence and theory around AD research community focus on plaques. You also know or should know that there is good research being done on competing theories---ones treating the plaques, perhaps not as epiphenomena, but as secondary to a more subtle primary mechanism. If you think that more research dollars should go to the second group (and I would likely agree), cool. But if you seriously think that the hypothesized connection between Aβ plaques and AD is based on an easily dismissed "logical error", you're not competent to participate in that discussion.

The entire field of Metascience (see https://en.wikipedia.org/wiki/Metascience for a basic introduction) exists to examine the question of whether people who think they are doing science are really doing Cargo Cult science.

Their conclusion? Most ARE doing Cargo Cult Science. Particularly those working in psychology and medical research.

Let that sink in. Because it really, really matters. And given that it is true, perhaps you should stop objecting to the characterization and start thinking about what would be required for that to no longer be true.

Now you said, If you think that more research dollars should go to the second group (and I would likely agree), cool. Well, that is a mild version of what I am saying. I am saying that the decades long refusal of the field to fund or allow publication of any alternative lines of research is a genocidal crime. I mean that quite literally. They wasted billions of dollars entrusted to them and caused millions of unnecessary deaths.

This is not to say that amyloid research should not have been pursued. It was the leading hypothesis for good reason. But doctors have learned the hard way the importance of giving both a diagnosis and a differential. And then to look for evidence for the differential.

Researchers need to be better at doing the same thing. And as long as they are apathetic or, as you have been, openly hostile to it, they will continue doing Cargo Cult science. No matter how impressive the procedure looks, they are failing to do the basic things that are likely to make it work.

Straw man. I made no assertion about the the prevalence of cargo cult science in Medicine or any other field. I, too, have read Ioannidis. And Smolin. And Goldacre. And Kuhn. And Popper(well, some Popper. His writing is pretty dense.) Hell, I've even read Robert Whitaker. My point continues to be that mere use of correlation as evidence of cum hoc ergo pro is neither necessary nor sufficient for a charge of bad (or cargo cult) science because cum hoc ergo pro is a logical fallacy.

Your reference to Metascience is a red herring. Ionnides didn't conclude that basic logic errors were the problem: his focus was on reproducibility and statistical errors as well as problems in experimental design (probably publication of negative results as well, I can't remember). Further, 'Why Most Published Research Findings Are False' was published fifteen years ago. It's mainstream science now, and researchers are entirely aware of it. Feel free to charge (with evidence) that it's being ignored in experimental design, but it's certainly neither controversial nor obscure. It shows up in High School science curricula. Hell, I've used it in freshman college maths.

Feynman and Ionnides focus on experimental design, evidence selection, researcher bias or incompetence in selecting results, and publication of results. They don't say that you can't use correlation as evidence: the problem is not that there's too much evidence being accepted, but that there's too little.

Had you opened with - "they're ignoring evidence they're wrong" or - "this paper has twice failed in replication attempts" or - "the study has too few subjects / no control group / too much noise to accurately measure these weak results" or even went all Smolin with - "the entire research program is based on inadequate and uncritical investigations into a poorly defined and contradictorily applied hypothesis",

I'd be merrily mashing the up arrow. But you opened with "The amyloid beta hypothesis is the result of a failure to understand that correlation does not imply causation." You're asserting that hundreds of thousands of qualified medical researchers don't understand a fundamental principle of the scientific method.

Somewhat amusingly, the published results that prompted this particular discussion is good evidence against the plaque hypothesis (well, it will be after replication and a supporting metastudy). From a well-regarded researcher. I guess they're not all participating in the genocide.

I'm not "hostile" to the idea that AD research might have been misguided or even criminally wasteful of resources. I'm hostile to the assertion that the very close correlation between plaques and AD should not be guiding research and that it is clearly not only wrong but can be easily so demonstrated.

Here is the thing. The message from medicine has always been, "Well yes, there were problems in the past, but we do better now." And then they don't. A new round of problems shows up, gets a new round of criticism, a new round of, "things are better" happens, and the same old thing crops up. This pattern has literally held for centuries. Why should we believe them now?

Here are some examples. When I was growing up in the 1980s, we were taught ancient history like how much doctors resisted the idea of handwashing. Those who were curious could learn about the struggle to get double blind studies, and past disasters like the adoption of frontal lobotomies, or the development of hyper-radical mastectomies based on the theory that cancer spread slowly through tissues. (Breast cancer cells actually get metastasized through the blood.) I was so curious, and learned that we do better now.

Except that in the 1990s we found out that millions had had false memories implanted through bad therapy, and the idea of evidence based medicine became mainstream, clarifying how much better we didn't actually do in the 1980s. Under Bill Clinton we even briefly had a committee created to identify the most egregiously ineffective treatments to save costs. That committee's first recommendation was that except in obvious critical trauma cases, back surgery be a last resort to be tried after every other treatment option. The reason why was that it was the most expensive option, was the least likely to be effective, and long-term outcomes were worst.

Sadly, back surgeons were more effective at politics than at treating back pain. (They are also very profitable for hospitals.) Therefore that was also the committee's last report, and it sank without a trace. According to https://www.beckersspine.com/spine/item/45273-surgeons-perfo... reports we do roughlyt 1.6 million back surgeries per year, despite the fact that we have long had evidence indicating that we should do fewer of them. Let's state that a different way. Each year we spend billions of dollars on making the lives of over a million people worse because there is no willingness to act on data that has been available for decades.

THAT is the state of modern medicine.

Oh, and the much hyped move towards evidence based medicine? Every doctor will tell you that we now engage in evidence based medicine, the bad old days are gone. But estimates are that only half of medical procedures that we do have evidence of effectiveness. See https://www.vox.com/the-big-idea/2017/12/28/16823266/medical... for a popular article on that. It also shows that back surgeons have a lot of company. I guess things are better. We have some idea of how much current medical practice is useless..we just aren't doing anything about it.

Now do you see why I don't accept that things have fundamentally gotten better?

Now back to Alzheimer's.

Until less than 3 years ago, the field was openly hostile to research and researchers that didn't accept the amyloid hypothesis. This changed not because the field internally realized that it was doing anything wrong, but because of criticism received from outside the field. In Nature no less. So despite their theoretical awareness of how to do science, they failed to connect that to their own practice.

How many other fields are making similar mistakes but simply haven't been called out on it? My best estimate is, "A lot."

And yes, I did overstate my case in the passage you quoted. A more accurate statement is, "The unquestioned dominance of the amyloid beta hypothesis is the result of a failure to understand that correlation does not imply causation." As I have said repeatedly, there was nothing bad about following up on lines of evidence suggesting that amyloid beta was a good lead. The problem was rejecting all research into ideas that might question the favorite theory.

You know some other examples of where medicine went astray because people involved in a field rejected research into ideas that might question practitioners favored theory? Let's see. Frontal lobotomies, radical mastectomies, repressed memory therapy, back surgery, knee surgery, and so on.

At what point in this list should we start to suspect that there might be a pattern?

Well, I didn't avoid the flames entirely. I singed myself a bit with the second article I linked. You might enjoy it if you haven't already read it

I have read it, but I still enjoyed re-reading it.

My interest in the subject was sparked some 20 years ago when I ran across an article about how people who question the amyloid hypothesis were being shut out of academia. I read up on it, drew the conclusions that I described above, then filed it away.

My gratification at having come to the right conclusion then doesn't match my annoyance that collective groupthink in the field has cost us collectively so much.

Don't worry, we have now identified tau tangles. This must be the root cause, since it is unfathomable that a complex neurodegenerative disease could have complex causes. Now that we have identified [insert misfolded protein candidate here], we can finally create a mouse models with the protein broken, then fix the protein, and then announce a cure. /s

Just so that this comment has something constructive in it - I'll add that I think it's important to understand that scientists are human and are also drawn by the allure of oversimple explanations for complex issues. For alzheimers, of the cause was a single protein or gene or something we already understand, we would have found it already. This isn't some ultra rare disease where nobody is looking. Its hard to say "I dont know," but that's the truth. The answer to this is going to come out of some obscure basic research that doesn't even seem obviously tied to neurology yet - and whatever team is working on it is going to have no idea how important their work is and is still going to get a great "I told you so!" moment out of it with regards to all the people asking them if their research has any practical purpose right now.

Mistyping "HIV-1" when it should have been "HSV-1" is a pretty unfortunate typo. Otherwise, great post.