5 ms·

Jülich quantum computer solves protein puzzle

48 points by donutloop 4y ago

fsh 4y ago
More accurate title: "Jülich analog computer solves toy version of protein puzzle".
- smitty1e 4y ago
  If these are toys, then my effort is dryer lint. Mock not the steps to answering a tough, heuristic challenge.
  - marginalia_nu 4y ago
    Quantum computation is still very, very long away from any sort of practicality. Finding another class of problems it can hypothetically solve given a spherical cow is neat, but doesn't really bring us any closer to having such a machine.
- gilleain 4y ago
  Specifically, the toy model they are using is mentioned as an 'HP' (hydrophobic-polar) lattice model: https://en.wikipedia.org/wiki/Hydrophobic-polar_protein_folding_model https://en.wikipedia.org/wiki/Hydrophobic-polar_protein_fold... It's often drawn a little like a go board, with black and white circles on a grid. The intuition is that the model tries to capture the essence of a protein chain as a sequence of hydrophobic ('water-fearing') and polar (water 'soluble', in some sense) residues. The goal is to get a folded chain, with a hydrophobic core and a polar outside - roughly imitating the topology of a real protein. Obviously, you could just move the chain around at random, at each step calculating the energy of the conformation, and accept the move if it lowers the energy. This energy is calculated (I think) by giving one value to H-H neighbours, another to H-P neighbours, and another to H-H neighbours. At least, that's my understanding of HP models from a little interaction with the idea many years ago!
mkesper 4y ago
TLDR: Researchers proved viability of quantum computers for non-trivial research questions in their field (protein research). While working with simplified models and thus being still years away from the complexity of problems solved right now with classical supercomputers, results are promising. Instead of running Monte-Carlo simulations and achieving about 80% accuracy (for even small amino acid chains), it's easy to achieve 100% with the quantum annealer. Still two to three generations of quantum computers expected to be needed for most research problems like drug research.
kensai 4y ago
The first such device out of North America. Why so few?
- cwillu 4y ago
  The short answer is because D-Wave machines are useless for actual problems. https://www.forbes.com/sites/alexknapp/2011/05/24/q-and-a-with-prof-scott-aaronson-on-d-waves-quantum-computer/ https://www.forbes.com/sites/alexknapp/2011/05/24/q-and-a-wi... goes into the details, very little has changed in ten years. “One thing this means, in particular, is that any claims by D-Wave that the practical value of quantum annealing has already been demonstrated need to be taken with a huge grain of salt.” https://www.scientificamerican.com/article/d-wave-scientists-line-up-for-world-rsquo-s-most-controversial-quantum-computer/ https://www.scientificamerican.com/article/d-wave-scientists... “D-wave machines are a long way from showing the exponential speed increase over classical computers that their advocates hope to see. But in a paper posted on January 17 and not yet peer-reviewed, a D-Wave team claimed the 2000Q could find solutions up to 2,600 times faster than any known classical algorithm (J. King et al. Preprint at arXiv https://arxiv.org/abs/1701.04579 https://arxiv.org/abs/1701.04579; 2017). Now the onus will be on sceptics to try to find a faster classical algorithm. “All I know is that, in the now two or three previous cases where we were in this same situation, it did turn out that a different classical solver eliminated the claimed gap,” says Scott Aaronson, a computer scientist at the University of Texas at Austin.”

48 points by donutloop 4y ago

fsh 4y ago

More accurate title: "Jülich analog computer solves toy version of protein puzzle".

smitty1e 4y ago
If these are toys, then my effort is dryer lint. Mock not the steps to answering a tough, heuristic challenge.
- marginalia_nu 4y ago
  Quantum computation is still very, very long away from any sort of practicality. Finding another class of problems it can hypothetically solve given a spherical cow is neat, but doesn't really bring us any closer to having such a machine.
gilleain 4y ago
Specifically, the toy model they are using is mentioned as an 'HP' (hydrophobic-polar) lattice model: https://en.wikipedia.org/wiki/Hydrophobic-polar_protein_folding_model https://en.wikipedia.org/wiki/Hydrophobic-polar_protein_fold... It's often drawn a little like a go board, with black and white circles on a grid. The intuition is that the model tries to capture the essence of a protein chain as a sequence of hydrophobic ('water-fearing') and polar (water 'soluble', in some sense) residues. The goal is to get a folded chain, with a hydrophobic core and a polar outside - roughly imitating the topology of a real protein. Obviously, you could just move the chain around at random, at each step calculating the energy of the conformation, and accept the move if it lowers the energy. This energy is calculated (I think) by giving one value to H-H neighbours, another to H-P neighbours, and another to H-H neighbours. At least, that's my understanding of HP models from a little interaction with the idea many years ago!

mkesper 4y ago

TLDR: Researchers proved viability of quantum computers for non-trivial research questions in their field (protein research). While working with simplified models and thus being still years away from the complexity of problems solved right now with classical supercomputers, results are promising. Instead of running Monte-Carlo simulations and achieving about 80% accuracy (for even small amino acid chains), it's easy to achieve 100% with the quantum annealer. Still two to three generations of quantum computers expected to be needed for most research problems like drug research.

kensai 4y ago

The first such device out of North America. Why so few?

cwillu 4y ago
The short answer is because D-Wave machines are useless for actual problems. https://www.forbes.com/sites/alexknapp/2011/05/24/q-and-a-with-prof-scott-aaronson-on-d-waves-quantum-computer/ https://www.forbes.com/sites/alexknapp/2011/05/24/q-and-a-wi... goes into the details, very little has changed in ten years. “One thing this means, in particular, is that any claims by D-Wave that the practical value of quantum annealing has already been demonstrated need to be taken with a huge grain of salt.” https://www.scientificamerican.com/article/d-wave-scientists-line-up-for-world-rsquo-s-most-controversial-quantum-computer/ https://www.scientificamerican.com/article/d-wave-scientists... “D-wave machines are a long way from showing the exponential speed increase over classical computers that their advocates hope to see. But in a paper posted on January 17 and not yet peer-reviewed, a D-Wave team claimed the 2000Q could find solutions up to 2,600 times faster than any known classical algorithm (J. King et al. Preprint at arXiv https://arxiv.org/abs/1701.04579 https://arxiv.org/abs/1701.04579; 2017). Now the onus will be on sceptics to try to find a faster classical algorithm. “All I know is that, in the now two or three previous cases where we were in this same situation, it did turn out that a different classical solver eliminated the claimed gap,” says Scott Aaronson, a computer scientist at the University of Texas at Austin.”