Slashdot Mirror
The Birth of Quantum Biology
Roland Piquepaille writes "Just when you finally have grasped the concept of quantum mechanics, it's time to wake up and to see the arrival of a nascent field named quantum biology. This is the scientific study of biological processes in terms of quantum mechanics and it uses today's high-performance computers to precisely model these processes. And this is what researchers at Rensselaer Polytechnic Institute (RPI) are doing, using powerful computer models to reveal biological mechanisms. Right now, they're working on a "nanoswitch" that might be used for a variety of applications, such as targeted drug delivery to sensors."
A quick search turned up an article from 1966 which suggests quantum tunneling in a protein, so the idea of quantum mechanics in biology isn't all that new (and probably predates the article). Disclaimer: I've only read the abstract, I don't do research in that area, those without a university hookup might not get to read it even if they really wanted to.
Most of those models have used classical mechanics with really primitive ball and spring models and a bunch of ad hoc rules for bonding.
Doesn't it make you feel good to know that our freedoms are protected by politicans, lawyers and journalists.
I agree on the overhype. What the article fails to properly elucidate is that this is a common expansion of existing molecular modeling techniques. All modern molecular modeling simulations are based on equations of force and motion experienced by the individual atoms. Even with "simple" interactions such as electrostatics, the equations are often rendered into power series approximations of the more complicated higher order equations. This makes it easier to do computationally intensive calculations by dropping the latter parts of the power series. In addition, some forces just are not modeled. These decisions are based on the computational expense of evaluating each of the force equations. As computers get more powerful, more of these dropped factors can be added back into the set of field equations.
Certainly there are difficulties in optimizing the mathematics to add new equations to the field set, as would be the case for modeling proton tunneling or sharing. A field that takes these factors into account is a great step forward. One of the biggest problems with modeling of biochemical systems is the lack of accurate accounting of hydrogen bond interactions. Some molecular fields are better than others, but most rely on a fudging of partial charge electrostatics by weighting interactions that have the features of a hydrogen bond (angle, distance, components). Hydrogen bonds are not easy to model, because they are a strange partial bond between covalent and ionic bonds.
That the researchers of this linked article have begun to include factors that could account for more computationally intensive quantum mechanic evaluations in their molecular modeling field is an exciting step for biophysical computation. It is not, however, a groundbreaking combination of quantum physics and biology like the article and slashdot abstract/title suggest. These fields already lie within a shared spectrum, and the connections outlined in this article are not a novel discovery of connection, but a novel utility of an already established connection.
alohapsilo
It sounds like plain biochemistry given a new window dressing. Not exactly ... there's actually something new here (2-3 years old in fact).
There are 3 "levels" of Computational chemistry :
- ab-initio method : a resolution of the Schrödinger equation for the studied system with only a few approximations mandatory to solve problem more complex than the hydrogen atom. This method is fairly demanding on number crunching power and is applied on models of hundreds of atoms.
- semi-empirical method : this method is based on the Hartree-Fock method used in ab-initio with the inclusion of approximation issued from experimental data. This method is applied on models of thousands of atoms.
- molecular mechanics : this method uses drastic approximations in order to obtain a Newtonian system (linear system of equations). The major drawback of this method is that it does not allow chemical reactions, but conformational changes may be studied in solvated systems. This method may be used to model system of 100.000s atoms. This method was mainly developed to study biological systems (a middle sized solvated protein gives a system of 100.000 atoms) because a lot of biological phenomenons may be explained conformational changes (life use the "soft" chemistry beautifully).
However some biological mechanisms still involve some chemical reactions, and to describe those using today's computing power some techniques were developed in order to mix molecular mechanics and semi-empirical methods. In order to observe a localized reaction, only a small subset of the system is solved using the semi-empirical method, the remaining is solved using molecular mechanics. This is what is presented in the article ...
This technique (QM/MM : quantum mechanics / molecular mechanics) was described and used as soon as 1995 but its use increased during the last 3 years due to the availability of the mandatory computing power ( http://scholar.google.com/scholar?q=qmmm&hl=en&lr= &btnG=Search ).
QMMM is not new, but its application field is widening ...
*nix is userfriendly
Being able to solve physics problems analytically depends a lot on what "symmetry". It's kind of a misleading term... basically what we physicists mean by symmetry is that a system is unchanged by changing the value of some property. So a square cut down the middle is symmetric in that both halves have the same shape. As a result we can solve problems involving the square without worrying about which half we're dealing with.
The "spherical cow" case is similar. A uniform-density sphere is great for ballistics problems because you can characterize it with only 2 parameters: its radius and its mass. If it's a realistic cow, it becomes a lot more complicated... its mass is distributed non-uniformly, and it's got a complicated shape (and it can move!!).
The real art in physics is figuring out when you can use approximations! If a cow is orbiting the moon, it's probably an excellent approximation to treat it as a sphere in order to determine its orbit. But if a cow is dropped off a cliff, it's not such a good approximation... since its air resistance will depend a lot on its shape.
My bicyles
But when I realised how hard quantum simulations could be it started seeming reasonable again. Quantum simulations aren't just an order of magnitude more difficult. The order of magnitude of difficulty can itself be an order of magnitude bigger!
Doesn't it make you feel good to know that our freedoms are protected by politicans, lawyers and journalists.
Well, physics is much more than freshman ballistics problems, but you're correct in that the complexity becomes significantly more difficult. Eg, in elementary quantum mechanics one can build a 'Hamiltonian' for any system, and usually you approximate things such as excluding the Coulomb force between every set of electrons, and that neutrons, electrons, and protons act as tiny magnets so they interact that way, and that there are spin-relational effcts, etc. Each of these adds terms to the Hamiltonian, but usually there's a convergence as the correction terms are smaller and smaller and can be neglected. Actually, that's why QED is so easy but QCD gets harder, because secondary and higher interactions in QED have decreasing significance but no so much in QCD where things diverge.
So in some sense you know the basic 'laws' of the universe, and right now we have pretty reasonable understanding of most things, neglecting large scales (dark matter, dark energy) and large energies (Higgs boson, gravitons, etc). But for stuff within our local spheres of observation, we have basic laws that account for most things we can see, so we should theoretically be able to model anything in this frame. The problem is that it becomes super complex very quickly.
Okay, so why there are so few solvable problems is mathematical. Eg, in the hydrogen atom, we can easily solve the differential equation that comes from the Schrodinger equation. Ie, you write the kinetic energy as T=p^2/2m, you write the potential energy as U=-e^2/r, giving a total energy of E=p^2/2m - e^2/r. You should recognize this as the standard kinetic energy written using momentum instead of velocity, and the Coulomb potential energy between the electron and proton. The system is an electron orbiting a proton, and in the center-of-mass units r is the distance between the two, and m is the reduced mass, which is fairly close to the electron. This is all well and good, and when put into the realm of quantum mechanics, r and p go from being canonical coordinates to being canonical operators. When put into the position basis, the p operator acts as a derivative of the r coordinate, and this yields a differential equation that must be solved to give the eigenstates of the solution. The system is spherically-symmetric which makes things much easier, and after solving the three-dimensional 2nd-order differential equations you get the solutions of atomic orbitals that you probably studied about in high-school chemistry class.
Now this is the 'simple' system. When you start adding relativistic corrections to that kinetic energy and when you add the interaction of the electron's magnetic moment interacting with the magnetic field creating as it orbits the proton, this yields the fine structure. You can also add in the spin-spin interaction between the magnetic moment of the electron and the proton, which gives the hyperfine interaction. Each of these things makes the differential equations MUCH harder to solve, and at some point we just don't mathematically know how to solve these complex systems of equations. Helium atom gets much harder because there are now two position coordinates of each atom, and an extra Coulomb interaction term. This is a quantum three-body problem, and even in classical mechanics the three-body problem cannot be solved in general. Ie, there is no KNOWN exact solution for any three bodies.
Anyway, you can see where this is going. But while we cannot know exact solutions, we can approximate them numerically to arbitrarily-small precision (at least with classical mechanics where there is no uncertainty principle). This is where the shiny computers come in. We can model easily how 10 bodies orbit around the sun AND interact with each other, but to get a general algebraic solution of them for any point in time, we cannot do.
make world, not war
It's more accurate than what most people believe because they spend so much time jerking each other off.
That's the big variational transistion state theory guy, right? Pay special attention to the details of how those potential energy surfaces are contructed - especially from the groups at that university. All of their QM/MM results match experiment almost perfectly (because after doing a QM/MM simulation they either "correct" the resulting "potential of mean force" curves or "correct" the effective PES obtained from simulation such the corrections cause the correct answer to be obtained.) In other cases, they parameterize the models until they get the answer that they wanted to get.
Wanted to get X, but get Y. OK. Let Z=(X-Y) and call Z a correction. In most any context, it's called cheating, but for QM/MM, it's the norm.