Cancer Isn’t a Logic Problem (2017)
cancer.nautil.us
cancer.nautil.us
He knows that this approach won't work before it has been attempted? I'll listen to criticism when it is followed by a suggestion for a better approach. And vacuous statements like "step back and look a the bigger picture" don't count.
>our fundamental constitution as multicellular organisms, our limited lifespan, the epidemiology of the aging population, socioeconomics and the future of society.
Like, hey man, you do realize that the killing or reprogramming of cancer cells goes far beyond the molecular technicalities of actually doing it?
Yup. It’s a pretty vacuous critism.
It cool to have some sed, but we may not even need that. We just need the equivalent of basic modern unix tools, like find, grep and rm, that work reliably, with good specificity and sensitivity.
Yet basic unix tools in medicine are decades away. Maybe it will be nanotechnology, maybe it will be more hacking around current solution.
Still, the problem is just:
find /body -type cancer_12BA |grep -v cancer_12BA_market_but_usefull_stemcell |xargs rm
The solution is technological, not philosophical.
Frankly, I don't see how biochemists today could possibly keep up with the advancements in AI and big data that have happened in just the last few years (and vice versa).
I'd like to see everyone take a step back and look at treating cancer as solving a large matrix where no single cause leads to an outcome. We need genetic algorithms to quickly find local maximums and minimums, we need machine learning to see correlations between seemingly unrelated findings, we need better standards for getting longterm studies into formats that are easily grokked by AI.
These are all straightforward, but I sense that there may be some resistance from the medical community in opening up the research similar to something like Folding@home. And in fairness, medical researchers are doing the heavy lifting. The first step probably has to come from hackers in the data science community, working with existing public research, and they're going to need real funding (which may not come until after the first of many cures is spotted).
Cells, however, are not files, so I would be careful extending the analogy. There is no "find" command for the same reason you can't run "find" on your physical bookshelf--objects in 3d space are not indexable files on a hard drive.
Perhaps we'll eventually have microscopic robots that can search out specific cells, in the same way Amazon uses robots to retrieve items in their warehouses. But that's a very different problem space, where the focus is largely mechanical rather than programmatic.
In particular, the quote is not meant as some sort of claim about the metaphysics of cancer or anything ridiculous like that. Rather, it's a (perhaps not so eloquently stated) way of pointing out that the mathematical and biological models underlying the current (and again, concrete) moonshot attempts fail to account for certain important issues and that this strongly suggests that those models are not going to result in cures.
The rest of the article summarizes some of Brivik's warrants, which are well thought-out. The claim is that the models currently used are fundamentally incapable of explaining how certain environmental factors effect cancer and the rest of the body, and that these environmental factors are important for understanding how to "solve" cancer because cancer evolves over time in often unexpected ways. The article points out, for example, that there are modeling assumptions baked into Microsoft's approach that are empirically falsifiable.
Finally, I'll just re-iterate that the argument made in the article is made in the context of current proposed/ongoing moonshots, so various comments about nanobots at the like are totally off-topic.
The article is not about science fiction. The topic is not whether maybe one day in the far future we might have enough knowledge about cancer cells to take a purely mechanistic approach toward curing cancer. Whether, the article is about the allocation problem in real science, today, and makes an argument against the efficacy of the approach taken by some current, concrete, well-funded moonshots.
Also, for those here: checkpoint inhibitor therapies are a "better approach" that incorporate a systemic activation of a patient's immune system to kill cancer, and they really, really work (see: Nobel Prize this year to Jim Allison).
As someone who also works for a company that makes a checkpoint inhibitor, one of our chief efforts is to better understand what makes them work for some patients and not others. Not only would this knowledge serve the lucky patients by tuning their therapy for better efficacy and lesser side effects, but would help us know more about how CPIs fail, and ideally, to enable their use in the many, as yet, unlucky patients.
As for the question of whether nanobots can completely cure cancer, that's scifi. Actually identifying tumors and removing them is a hard classification problem followed by a careful surgery. Generally this is done using advanced imaging, radiologists who interpret the images, guiding surgeons or doctors to either do tumor resections, or select chemotherapy. Nanobots are just a tiny version of the tumor resection, with the added problem that nanobots don't exist, they don;t' have the computational power to discriminate invasive tumors (at least, the current generation of "robot pills don't).
So let's focus on what is tangible in the near future: additional use of machine learning based on large training sets collected across hundreds of studies. That is in fact very likely to counter the author's hypothesis.
If it is physically feasible, which it is given we have immune cells and they are just evolution-created versions of these nanobots, then it is just an engineering problem and we humans are pretty good at engineering things.
Things will keep moving forward; macro approaches such as your suggestions will work insofar as we can find the cells. In the end it will probably be a mixture of re-engineering our immune cells to better identify and protect against cancer, macro approaches such as the ones you mentioned, and more.
It is a hard problem, of course, otherwise we'd have it solved by now, but we've solved hard problems before. And with AI it will be easier and easier to do so.
We’re not yet at nanoscale manufacturing in one of the few domains with the scale and capital to fund its development: microprocessor manufacturing.
Re-defining biology as “nanobots” renders the latter word useless.
Paper planes and fighter jets run can be modelled by the same physics. That doesn’t mean they’re the same thing. “Nanobots” implies greater degrees of freedom and function than purely biological techniques can manage. (For example, by incorporating digital circuitry and novel materials into the mix.)
As I said, they can include more types of atoms, but that adds little as life already makes use of a wide variety. It’s basic capacity that’s the issue not micro optimizations. Further, depend on a non organic atom and you can’t self replicate without being fed it.
At the smallest scale you can’t for example take a picture, you need to detect stuff by touch and respond before bouncing off while traveling at 50+MPH. Make a vast structure like a cell and movement and energy become major issues.
“Digital” means a logic model. Not a physical technology. Atomic-scale digital circuitry—completely foreign to life as we know it—is certainly plausible.
Your claim is that anything that can be done at the nanoscale can be done by “editing DNA to make a [protein].” I’m saying the latter is a subset of the former. Given all we have, at this point, is theory, it’s a reasonable point on which to disagree.
That’s not to say proteins are as efficient as possible. Just that they are fairly close to the limit of what’s possible at 1 to 100nm. Now start talking micrometers‘s and I can see some things changing.
Science Fiction vastly overstated what physical things at this scales can do. I can describe anti gravity fairly easily, that does not let people build something.
Nanobots are not necessarily just tiny tumor resectors, they could be implemented in dozens of possible ways with future technology (using some form of machine learning).
The author is realistic. What I see among young people today is technology worship. It's like technology has a mystical power over you guys and you think that it will solve every problem known to man. Technology has limits just as physics has limits, but you guys bend all problems toward the narrative of being solve-able by Data science, machine learning or nano-machines.
I hate to break it to you. Not only will many problems remain unsolved. The future will proceed in an unpredictable direction.
>Technology is just applied and abstracted physics.
Like thats it. Technology is so simple. It's just physics. I'm arguing with your attitude and your technology worship. AS if nanotechnology is right around the corner and will cure cancer 100%. Cancer is not a trivial problem and likely a problem that will not be solved within our lifetimes.
All the article is saying is that data science, machine learning and all these buzz words popping out of silicon valley aren't the answer to cancer. But we have people like you heralding machine learning and nano-machines as the answer to problems such as world peace. Open your eyes.
The same thing that currently prevents nanobots (your immune cells) for eradicating the problem of cancer. So, why don't we figure that out?
Everybody in tech not in biology needs to realize that we HAVE nanobots--they're called cells or bacteria or viruses (also called minimum viable product). And, most of the time, those nanobots do a good job. However, sometimes they have bugs that get exploited and we need to be able to help out when that happens.
Cancer is definitely shapeshifting and it's a constant battle on all fronts.
One drug may work now (eg. anti folates) but then remission fails and it's back to the fight.
The primary tumor may go into semi-remission but mestasis may happen then you have to treat that instead.
The tumors suck so much energy from the person's body so it can attack on that front too.
I think immunotherapy is the future and am really sad that it is so expensive. It's like Elysium, Keytruda is ~$400k it's sad how costly these drugs are.
I think diet, environment, and habit are the things that have the most effect over time because it's really just your cells multiplying in an uncontrollable rate and the body cannot stop the function or the garbage collectors are unable to make the cell perform aptosis.
There are many things that go in and out of your body over time that will help increase or decrease the amount of these types of cells that are produced.
https://en.wikipedia.org/wiki/George_Dantzig#Mathematical_st...
that is, "not at all in the usual case"
XKCD isn't criticizing the positive mindset of naive-but-gifted researchers; it's criticizing the attitude of experts in one field that their expertise necessarily gives them unique and powerful insights into solving problems in another field.
Perhaps a better counter-anecdote would be all of the mathematicians in the 1950s - 1980s who solved hard program design problems with little or no experience actually programming computers. That really was a counter-example to the XKCD mentality: a bit of mathematics was far more useful than knowing the instruction set of the latest IBM mainframe...
I would argue that one of the biggest barriers to fighting climate change is the blind belief in technology as the solution to any problem facing humanity. In many ways it is a cult. But in order to fight climate change we might need to turn to simpler, more low-tech solutions.
I mean, we're spending all that effort flying to the moon and trying to program cancer cells, while all we really should do to fight cancer is clean up the air we breathe, stop spraying pesticide on our food and lead less stressful lives. Just saying...
And btw., how would you clean up the world?
Insofar as anything needs hacking, it is this wasteful focus in the research community. The research community needs a dramatic refocusing onto mechanisms that cannot be evaded and which occur in most or all cancers. The canonical example is to interfere in telomerase and ALT telomere lengthening. That can't be evaded, and all cancers do it. Shut these things down and cancer stops. Any cancer, any type, any stage, using exactly the same therapy. Yet the amount of work in this area is tiny in comparison to projects that verge on personalized medicine per tumor.
The problem with this is that if something cannot be evaded(is probably "core" to the cell) or occurs in most cancers, it also most likely occurs in healthy cells as well. 10000s of drugs are made each year which kill cancer cells, but the problem is that they are not specific to the cancer and kill everything else too. The holy grail in cancer drug development is to develop a drug that ONLY kills cancer cells and leaves everything else alone. We have yet to make one of these.
>The canonical example is to interfere in telomerase and ALT telomere lengthening.
Cancer cells do lots of random things. After all, they are just a collection of "bugs" that occurred randomly in such a way that they present symptoms we recognize as cancer. Just because they express telomerase doesn't mean its important. Im sure they express lots of random other things too. "Obvious" solutions like this, have been tested by 100 people by now, and since we haven't heard anything, we can safely assume it doesn't work.
Yeah, and so do adult stem cells. As long as your patient doesn't need intestinal walls, skin, or blood, this will be a totally successful treatment!
Have you entertained the hypothesis that people studying cancer might know more about biology than anti-aging cranks?