3 ms·

Before CRISPR came along, I remember talking to someone working in archaebacteria who went on and on about these arrays of sequences in archea that seemed reall

Before CRISPR came along, I remember talking to someone working in archaebacteria who went on and on about these arrays of sequences in archea that seemed really fascinating to him, but I never quite got the implication of what this would eventually develop into before it actually did become CRISPR (by other people working in the field). I remember ~20 years ago hearing about nanopore DNA sequencing, and wondering if it would ever work (it does!). I would have definitely funded both of these people, but I wouldn't necessarily have funded the projects. I knew they were brilliant, but I didn't know exactly where it would lead or if it would work. The key really is to break out of the current "big famous lab" approach, IMHO. Or at least have two tracks of funding. I'm not sure about how to fund people not projects in a systematic way, but we definitely need more random shots on goal and more stochastic exploration of the space of possible research areas.

mattkrause 5y ago
Why not both? It's astonishing that we have, essentially, a single "one-size-fits-all"[0] mechanism for funding biomedical research. We should have a bunch, ranging from "trust me, I'm a genius" to "you've never heard of me, but the data suggest this will work." There should be mechanisms where tons of preliminary data are required, but also mechanisms to generate that prelim data. There should be mechanisms for trainees, but also to keep experienced people in the field. [0] Essentially, the NIH's R01. Even that needs work because inflation has chipped away at what one modular budget can support.
- epistasis 5y ago
  Yes, I think I'm suggesting both! I also agree with how R01 has changed as they become more and more competitive with lower funding rates. There's also been a rise in "but science" projects in bio, starting with the human genome project, that mirrors some of the big science that happens in physics. So adding in that track of funding there are kind of three funding tracks that are necessary.
bsder 5y ago
> Before CRISPR came along, I remember talking to someone working in archaebacteria who went on and on about these arrays of sequences in archea that seemed really fascinating to him, but I never quite got the implication of what this would eventually develop into before it actually did become CRISPR (by other people working in the field). >I remember ~20 years ago hearing about nanopore DNA sequencing, and wondering if it would ever work (it does!). These two things are very different to my eyes. Nanopore sequencing seems obvious to me as an electrical engineer. The basic problem was noise and control--which you could probably solve with computes and Moore's law. You can fund it, but it feels like something you could leave to industry as it seems like it's engineering more than science. CRISPR is non-obvious. It's more in the realm of Taq polymerase--it's interesting but, so what? Well, Taq polymerase would become the keystone to PCR about 15 years later. I'm much more interested in figuring out how to fund the things that later become building blocks rather than pouring money on something that industry is likely to do anyway.
- zozbot234 5y ago
  > I'm much more interested in figuring out how to fund the things that later become building blocks rather than pouring money on something that industry is likely to do anyway. Industry will only ever fund things that are easy to monetize in the very short term. Translational research can be just as important as basic research, so both should be funded when they are plausibly in the public interest, regardless of any private returns.
  - ahartmetz 5y ago
    > Industry will only ever fund things that are easy to monetize in the very short term. Usually. ASML worked on something for two decades, and now, against all odds, EUV works :) But, of course, it is notable because it is far from the normal case. And it was, more or less, an engineering problem, which required some support from scientists.
- epistasis 5y ago
  Both of those issues: 1) slowing down the DNA, and 2) finding enough change in current to be able to distinguish bases, did not seem that they could work out. You really have to solve the first before you know if the second is solvable, but getting tiny molecules to behave as they flit through protein pores is not a trivial task, particularly around ~2000. The signal processing is still quite significant, and I hear that if you don't have a compute box that's set up correctly, you can't get data to and from the GPU quickly enough for base calling (no personal experience, just heard friends griping about early experiences). So even with todays compite tech, the signal processing is pretty heavy duty, 2000 era hardware and algorithms would have far far less accuracy. Which is all to say that yes, now it works, and my uncertainty was perhaps unwarranted. I'm super impressed that it works, and I've learned a ton of humility about guessing what other people can figure out how to do. David Deamer is a visionary, and all the people that worked so hard on it afterwards deserve major kudos.

mattkrause 5y ago

Why not both? It's astonishing that we have, essentially, a single "one-size-fits-all"[0] mechanism for funding biomedical research. We should have a bunch, ranging from "trust me, I'm a genius" to "you've never heard of me, but the data suggest this will work." There should be mechanisms where tons of preliminary data are required, but also mechanisms to generate that prelim data. There should be mechanisms for trainees, but also to keep experienced people in the field. [0] Essentially, the NIH's R01. Even that needs work because inflation has chipped away at what one modular budget can support.

epistasis 5y ago
Yes, I think I'm suggesting both! I also agree with how R01 has changed as they become more and more competitive with lower funding rates. There's also been a rise in "but science" projects in bio, starting with the human genome project, that mirrors some of the big science that happens in physics. So adding in that track of funding there are kind of three funding tracks that are necessary.

bsder 5y ago

> Before CRISPR came along, I remember talking to someone working in archaebacteria who went on and on about these arrays of sequences in archea that seemed really fascinating to him, but I never quite got the implication of what this would eventually develop into before it actually did become CRISPR (by other people working in the field). >I remember ~20 years ago hearing about nanopore DNA sequencing, and wondering if it would ever work (it does!). These two things are very different to my eyes. Nanopore sequencing seems obvious to me as an electrical engineer. The basic problem was noise and control--which you could probably solve with computes and Moore's law. You can fund it, but it feels like something you could leave to industry as it seems like it's engineering more than science. CRISPR is non-obvious. It's more in the realm of Taq polymerase--it's interesting but, so what? Well, Taq polymerase would become the keystone to PCR about 15 years later. I'm much more interested in figuring out how to fund the things that later become building blocks rather than pouring money on something that industry is likely to do anyway.

zozbot234 5y ago
> I'm much more interested in figuring out how to fund the things that later become building blocks rather than pouring money on something that industry is likely to do anyway. Industry will only ever fund things that are easy to monetize in the very short term. Translational research can be just as important as basic research, so both should be funded when they are plausibly in the public interest, regardless of any private returns.
- ahartmetz 5y ago
  > Industry will only ever fund things that are easy to monetize in the very short term. Usually. ASML worked on something for two decades, and now, against all odds, EUV works :) But, of course, it is notable because it is far from the normal case. And it was, more or less, an engineering problem, which required some support from scientists.
epistasis 5y ago
Both of those issues: 1) slowing down the DNA, and 2) finding enough change in current to be able to distinguish bases, did not seem that they could work out. You really have to solve the first before you know if the second is solvable, but getting tiny molecules to behave as they flit through protein pores is not a trivial task, particularly around ~2000. The signal processing is still quite significant, and I hear that if you don't have a compute box that's set up correctly, you can't get data to and from the GPU quickly enough for base calling (no personal experience, just heard friends griping about early experiences). So even with todays compite tech, the signal processing is pretty heavy duty, 2000 era hardware and algorithms would have far far less accuracy. Which is all to say that yes, now it works, and my uncertainty was perhaps unwarranted. I'm super impressed that it works, and I've learned a ton of humility about guessing what other people can figure out how to do. David Deamer is a visionary, and all the people that worked so hard on it afterwards deserve major kudos.