This is one key comment that fundamentally differentiates software/hardware from biotech: "Overall the power of recombinant technology and its rapid growth in the 1970s/1990s didn’t necessarily translate into success for new technologies." Powerful abstractions and pervasive feedback loops across the hardware and software stack is what fundamentally enabled rapid innovation in tech; those same ideas haven't developed even close to the same degree in the biotech space.
That's because we lack sufficiently precise computational models of biological systems. It is the "fusion power" problem of biotech. There has been some progress, but this isn't an easy problem to solve (just like controlled fusion).
One of the reasons why biotech has so low success is that we see it as kinda ok to unleash a half tested app onto people, basically most startup apps out there, while new drugs are extremely hard to get tested.
It's kinda understandable given that some drugs can have serious negative long term effects on people, whereas a buggy CRUD app can't really do much harm in comparison.
While the costs of a "buggy" drug are usually larger than the cost of a buggy app, the benefits of a "good" drug are usually larger than the benefit of yet another app to share pictures, or a uber but for X.
Just see how much it costs to save the economy just from the corona measures. The opposite of that is the benefit of a 1 year increase of people's healthy life spans, much more in fact as only parts of the economy were shut down.
modern pharma R&D is pretty much move fast and break things. All modern pharma firms spend enourmous effort to ensuring that errors are cought as early as possible (shift left) in R&D.
Hundreds of molecule candidates are discarded before reaching Phase 3 trials
I don't think this comparison holds up. In software, the reason to release something early is because it is hard to tell if there is demand for what you have built and getting feedback from the market is vital.
In pharma, I think the success criteria are much better understood, but it's just really hard to do what people want.
Not so say our regulatory system isn't a shitshow that needs reform, but I don't think there is the same benefit of releasing buggy versions.
This will change when we can spin up whole human bodies for experimentation and discard them at will.
We need to clone decephalized humans. Turn off neural plate development. Downregulate genes that cause brain development. Before we get this right we can supplement it with surgical techniques to ensure the clones never have brains and never become sentient. Hook the embryos up to artificial life support that innervate heart, lungs, and replace hormonal signalling.
We could have a facility full of live, brainless monoclonal humans. We could perform research that wasn't previously possible due to ethics and invasiveness.
As a side benefit, we now have a limitless blood and organ supply for transplantation. Ideally without any antigens that react with transplant patients, guaranteeing transplant recipients never need to take quality of life-altering immunosuppressants.
If we can treat real human bodies like lab rats, we'll conquer every cancer and have a real chance of doubling human healthspan.
We have to get out of "punch card" phase biology. Unfortunately, it'll require wading out into uncomfortable new frontiers that many won't be happy to broach. This requires that you accept that there is no soul, that humans are code, and that we are now the authors.
What makes you think that (1) that it’s even possible to create such beings, (2) that these would even be good disease models or (3) a lack of model organisms is the main factor holding back medical breakthroughs? None of these is at all obvious...
1. It'll be an incredibly difficult engineering task with many large hurdles. But it's possible, because life found its solution in us. We culture human cells all the time, just like bacteria or plants - this is just a higher structure. In fact, you can think of this entire process like growing humans as plants.
2. We spend a lot of time "humanizing" model organisms by creating chimeras. But even then the results are frequently not portable. How many animal studies fail to make the jump to advanced clinical trials?
3. A human is the system you want to study. It has the same genome, the same transcriptome, the same proteome, the same metabolome, and the same kinetics and thermodynamics. Animal models are 17th century dentistry in comparison.
Placebo research and documented cases of physiological diseases treated via mental health therapy puts in question the viability of using theoretical human-like creatures that possess no consciousness, and therefore no mental state to speak of.
To draw an analogy, this would introduce such a huge gap between “staging” and “prod” environments that the effort and the ethical risk of ending up with “staging” entities possessing any degree of human consciousness cannot even remotely be worth it. (I’m somewhat alarmed that I feel the need to spell that out.)
> Placebo research and documented cases of physiological diseases treated via mental health therapy
How is cancer something you can fix with mental health therapy? Tell them to go home and feel better, that it'll all be over soon?
What about a host of other genetic and immune disorders, many of which can be fatal?
I don't mean to come across as harsh or mocking, but there are a host of diseases where your solution does not work.
> I’m somewhat alarmed that I feel the need to spell that out
I understand your argument, but I reject it because I don't think we're seeing the same picture of where things now stand or could be. I'll counter with an analogy of my own.
Right now we're attempting to fix issues in prod by looking at another company's software, decompiling it, bit fiddling a bit, then trying to hot patch it into our own production instance. We run some smoke tests before releasing things, but these can't always identify long-term effects.
I want to have an actual honest to god dev and staging environment. Not just that, but a system that looks remarkably like prod that we can burn down and spin up on demand for each investigator trying to solve novel problems.
We can actually test at scale (and duration!) and pick apart the system to see that no part of it goes into failure mode or disease state.
Development goes much faster when you have an actual dev kit to run your code on or tear apart.
We don't have anything like that right now.
I get that the notion that human beings being equated to fungible software or disposable hunks of meat is uncomfortable. But the special ingredient in both you and me is consciousness, and this design ensures we never bring new consciousness to life. We turn off the gene expression that causes the brain to develop, and we can double down with mechanical amputation.
It sounds macabre, but so too did dissecting cadavers a century ago.
> Right now we're attempting to fix issues in prod by looking at another company's software, decompiling it, bit fiddling a bit, then trying to hot patch it into our own production instance.
The staging/prod analogy is flawed.
What we are doing is manipulating board wiring, while what we could do is further our understanding of and abilities to alter the OS and firmware running on it.
Very unfortunately, but understandably, scientifically oriented mindset tends to dismiss the latter route. After all, so far it seemingly rules out any notion of dev environment. Where is reproducibility, if all we can do is SSH into a snowflake live system (or more like XSS inject it, right)? A daunting limitation, but perhaps not impossible to overcome.
That is not to say the former has no place and your point is invalid. Still, when all you have is a hammer…
> Where is reproducibility, if all we can do is SSH into a snowflake live system
That's the beauty of this, though. You'd have monoclonal lines of decephalized humans. They're not snowflakes.
You'd have knockout lines for studying specific genes, ABO- lines for a blood supply, lines for all the major MHC/HLA for transplants...
If you have twenty identical deceph human clones, you can run your same experiment in duplicate and push for statistically meaningful results.
Over time we'd aim to reduce the costs of the bodies so that they are cheap to expend in individual experiments. That's when it becomes incredibly powerful, when the average researcher can afford these in their budget.
We might learn to control the rate of development. Quickly grow them in a way that doesn't interfere with the experimental results.
> That's the beauty of this, though. You'd have monoclonal lines of decephalized humans. They're not snowflakes.
Your approach does not duplicate the OS, only hardware wiring. Again, not to say fiddling with condensers and transistors can’t fix a misbehaving program, but it’s not always the best way to address the issue with fewest side-effects.
It's better to not think about biology as software vs hardware. Also, most diseases would be hardware defects in this metaphor, so why would fixing hardware not help? If you can treat a brainless person for the disease (we do that, e.g. if a mother is bearing a child, we might keep her alive even though she is brain dead), there shouldn't be any problem with using decephalized humans as models. Placebo effect works (e.g.) by tricking the immune system into action (doable in a decephalized human just as well), not by telling the disease to go away.
The assumption that most diseases are hardware is that, an assumption.
Some may be, sure.
But could it be that a blanket claim like that comes from us already knowing how to access the hardware, and being uncomfortable to wander out of that comfort zone?
Of course you can’t “tell” the disease to go away any more you can tell a bug to take a hike, but saying we can manipulate the hardware to trigger whatever immune, etc. responses just as well comes from that being the thing we can sort of do reliably at this point.
And even then, can we really?
Read about the phenomenon of cascading medical involvement during childbearing and birth: turns out “hardware manipulation”, which is very compelling to do, tends to have compounding hidden consequences, while the adjustments in the elusive “software” (inducing relaxed state etc.) somehow appear to yield straightforward results. Again, immediately life-threatening conditions aside.
Manipulating transistors and condensers (triggering an immune response, in your example) might just be a suboptimal, incomplete approach when compared to updating the firmware logic (making it so that the response is triggered when and how needed), or of course better yet both.
Our discomfort with the associated uncertainty, however, makes this an unfashionable direction of scientific exploration.
This is exactly where the software/hardware metaphor breaks. The software is the hardware, and the hardware is the software. The mind is - not influenced, it simply is exactly that - the processes in the brain, and the body acts on various kinds of signals sent by the brain and vice versa - and we should be perfectly able to simulate these signals if we can grow a brainless human body, and see the effects on the body. That would help us a lot in determining what needs to happen within the brain and in turn within the mind, for example, if said problem is psychosomatic or mentally fixable. However not everything is driven by the brain - and treatment for stuff like most types of cancer has nothing to do with the brain at all, same goes with vaccines development, stuff like Crohn's disease etc.
Let's say it's possible to trigger immune response by focusing really hard. Still we need the brainless human to determine which signals the brain needs to send before we start trying to get people to focus really hard.
There's absolutely nothing wrong with what I've proposed.
Typically only people that believe in a soul or that humans "should die" take issue.
Neither of these is correct in my world view. I can understand the religious perspective, but the romantic idea that limited live spans are good is incredibly twisted.
I think people should be immortal, but that's just not feasible. The next best thing is using all of the tools around us to extend our lives as much as possible.
I'm tired of biology being so punch-card slow while the compiler is literally staring us in the face. The human is the perfect model organism for everything we want to accomplish, and I've shown that we can ethically build upon it as a platform without harming anyone. Moreover, we can farm it. Never deal with blood or organ shortages, end immunosuppressants for many people, and potentially perform purely regenerative surgeries.
A human grown in a lab without a brain is a plant. Dogs, apes, and octopuses are more human than that. Before you fight me, go take on the many acts of animal cruelty in the world.
The people that would hold this back are Luddites.
This is a faster path to curing cancer. All fifty thousand types of it.
edit: Down voters, why not argue your point instead? Because from my perspective your views are a dated death wish.
THere definitely are a number of things "wrong" with what you've proposed. The first is that society simply wouldn't accept it; by the time people learned what you were doing, the government would make laws against it. People really are averse to growing brainless humans in vats for medical experimention and spare parts.
I'm not interested in "fighting" you,but the reality is that you're starting from a non-starter position that would never fly in the larger general public.
It's not worth debating, because you're absolutely, totally convinced that you're right. Most of us here agree that in principle it would be great to be able to run experiments that truly address human medical problems without ethical concern, but I think "decephalized humans" don't really qualify.
> non-starter position that would never fly in the larger general public
At one point in time the majority of the general public was against gay marriage. Two hundred years ago many were pro slavery. Everyone used to be against cutting up dead bodies so we could learn how they worked, so it was conducted in secrecy.
Most people are illiterate about the science behind this. The benefits are real and substantial. This isn't a money grab or vanity project - it pushes humanity forward a notch, and it should be done.
Anti-vax, anti-GMO, anti-scientific thought shouldn't prevent us from developing a tool that saves countless lives.
Where is the line? Is it okay to use human stem cells for experimentation? Is it okay to grow whole organs for replacement? Is it okay to grow muscles, ears, skin, eyes, bones...? Why is not okay to grow them all at once? I don't see any difference. All of this literally is "decephalized humans", just with more modification/specialization and a simpler process.
We're even growing brain tissue that exhibits actual human-like brain waves for experimentation - how is that okay with you!?
I'm an emergency physician and I think what you have proposed is unethical and immoral. Good luck getting any type of experimentation through the Institutional Review Boards(IRB), bioethicists, and scientists.
Call us Luddites but this type of science is not looked at kindly. There's no argument or debate to be had when what you are proposing is a red line for many and a non-starter.
The dev -> silicon -> user feedback loop is orders of magnitude faster, cheaper, and more scalable than one that involves wetlabs, clinical trials, regulatory bodies and interfacing with the medical-industrial complex.
Hi. This is pretty early for this, but I’m looking for some advice for starting a seed fund to advance early-stage startups in cancer immunotherapy. It’s not for me, but someone whom I did some machine learning work, and angel investment with. He’s one of the top names in the field from the early days of cloning to today, veteran of several biotech startups, and likely recent Nobel and Lasker winner advising. There’s a lot we don’t know about the financial side, and I’m trying to figure out whether this is feasible, and what the plan should be. Thanks!
Lots of variables but it probably is feasible depending on 4 factors:
how many viable cancer immunotherapy startups are there
2) Due diligence
how good are you at separating the wheat from the chaff
how much seed investment are they receiving vs how much they require, it does sound like a niche market
4) Strategy what are the timeframes involved, when do you get to “realize” your investment, riding them out to IPO?, etc
5) how will you accelerate their growth, contacts, help them with their own strategy etc...
Now these factors play into the success of the fund in their interactions. Example 1 and 2 are not ideal but 3 is a situation with little competition, you might be able to get better deals more equity.
Obviously a lot of domain knowledge is required but
if you are able to address these 5 things satisfactorily then you might easily find money chasing you.
After that it’s just coming up with a good investment structure (ex. certain international investors seek tax sheltering), some lawyers and choosing limited partners. Shouldn’t be too hard especially given the level and recognition this gentleman should command in his industry.
Noble goal, and I’d suggest the book Basic Immunology, but I’d also warn you, cancer might be a famously difficult disease, but the number of treatments is vast, thus it is difficult to market a new one, because oncologists are overwhelmed by options. There’s an idea for your ML buddy: just imagine rather than make new treatments, you could help docs pick existing treatments from the sea of complex options...
I personally read Janeway in 2014 when I learned about the problem and started helping with ML models. But the fund isn't mine; it will be run by the leader in the field. He's already heavily invested in a nonprofit that provides physicians and patients with information about leading-edge therapies: https://www.ucir.org/.
I am relatively new to the cancer IO space, but unless you discovered a new check-point idea (e.g., TIGIT) a lot of the ideas are flushed out already. If you're thinking cell-therapy, there are a number of autologous ideas already on CMC for p1 testing. If you're trying to make TILs angry, or modulate M1/M2 CD11b macrophages, anything that requires transcriptional re-programming, you have a lot of catching up to do.
Most of the stuff you mention here is more biological, which is outside of my full understanding as an ML hobbyist in the area (not my fund, just helping). There is some low-hanging fruit in the portfolio like an improved ligand for an existing checkpoint target. I worked mostly on predicting ligands to stimulate the HLA->APC->TCR->proliferation pathway, mostly using data from binding assays and HPLC/MS isolation. There is a lot just on the computational side that is required to apply well-known vaccination methods to cancer. For seed startups, I'd think it's more likely to find fundable companies making progress on improving the application of existing leading-edge methods than attempting entirely new approaches. CAR-T comes to mind as a prohibitively expensive therapy that invites approaches to reduce cost and improve access, which gives them a simpler path to commercialization than starting with the FDA from scratch. Again, not the expert, not my fund, so I can't comment on the need to catch up, but the people I'm trying to help are some of the luminaries in the field, so I don't think unfamiliarity is the major roadblock; it's more a matter of whether their time is best spent on this or on research.
If you are interested, I can connect you with a consultant in the field who works with VCs and young biotechs. Really great to work with and if he can't help you out, I am sure he knows someone who can help you.