Mechanical properties of protein fibers ======================================= We have been working with the experimental group of `Dr. Weisel `_ (University of Pennsylvania, School of Medicine) to resolve the structural details and the mechanism of force-induced unfolding of fibrinogen monomer :math:`Fb` and oligomers :math:`(Fb)_n` (see Figure). Although the physical properties of fibrin fibers, the major structural component of a blood clot, which control their function in hemostasis and wound healing, are have been fully characterized, the underlying mechanism of their force-driven elongation is not understood. We carry out computational studies of the mechanical properties of fibrin protofibrils on GPUs using the `SOP-GPU package `_, in order to characterize the micromechanics of fibrin at the monomer, oligomer, and fiber level. We use GPU-based computations to speedup the molecular simulations. For example, it takes ~15 days to obtain one trajectory of the mechanical unfolding for the oligomer :math:`(Fb)_3` of three :math:`Fb` repeats on a GPU GeForce GTX 480, using the SOP model implemented on a GPU (SOP-GPU). For comparison, it would take ~12 years of the wall-clock time to complete a single simulation run on a single CPU core of comparable level of technology, using the same SOP model. The results were published in `Structure (2010) `_. .. figure:: fibrinogen.png :align: center :figwidth: 50% **Figure:** Fibrinogen structure. *Left panel*: Crystal structure of fibrinogen. The central nodule, formed by the :math:`N`-terminal portions of all six chains, is connected to the distal :math:`\beta`- and :math:`\gamma`-nodules formed by the :math:`C`-terminal portions of the :math:`B\beta` and :math:`\gamma` chains, respectively, by triple-helical coiled-coils, each formed by the middle portions of the :math:`A\alpha`, :math:`B\beta` and :math:`\gamma` chains. *Right panel*: Same molecule as on top but with those regions that are not identified in the crystal structure, the interacting :math:`\alpha C`-domains attached to the molecule with the flexible :math:`\alpha C`-connectors, and the :math:`N`-terminal portions of the :math:`B\beta` chains forming the functional :math:`B\beta N`-domains. The funnel-shaped domain in the center contains fibrinopeptides A (FpA) and fibrinopeptides B (FpB); the :math:`\gamma N`-domain is located on the opposite side of the molecule (not shown). The individual domains of the D regions, i.e., A-domain, B-domain, and P-domain, are indicated only in one subunit of the molecule. The site "a" (hole "a") and site "b" (hole "b") in the P-domain of the :math:`\gamma`- and :math:`\beta`-nodules, respectively, are indicated by asterisks. (taken from *J.Thromb.Haemost* **7**: 355 (2009)). Forced unfolding *in silico* of the fibrinogen monomers and oligomers --------------------------------------------------------------------- Large-size protein systems unfold through the gradual detachment of two or several subdomain and/or through the simultaneous or sequential unraveling of the various secondary structure elements. This allows us to utilize coarse-grained descriptions of biomolecules to describe the global mechano-chemical unfolding reactions. We utilize the SOP-GPU package to carry out Langevin simulations of fibrinogen monomer :math:`Fb` (~2,000 residues) and dimer :math:`(Fb)_2` (~4,000 residues) using the experimental force-ramp conditions, :math:`f(t)=r_ft`, where :math:`r_f=\kappa \nu_f` is the force-loading rate. We use the experimentally relevant values of the pulling speed :math:`\nu_f=1` :math:`\mu m/s` and the cantilever spring constant :math:`\kappa=35` pN/nm. To obtain a single trajectory of unfolding for :math:`Fb`, the system dynamics should be propagated numerically over as many as :math:`5 \times 10^9` iterations (0.2 seconds of real time). To fully utilize the GPU computational resources, we employ the multiple-runs-per-GPU approach, which allows to run many trajectories for the system under the study concurrently on a single GPU. It takes ~17 day to generate 5 trajectories on a GPU GeForce GTX 295 (NVIDIA). For comparison, it would take ~18 months of wall-clock time to complete a single simulation run on a CPU of a similar level of technology. It turns out that the unfolding micromechanics and the corresponding dynamic signatures in the force spectra change dramatically with increased pulling speed (:math:`\nu_f=2.5` and :math:`25` :math:`\mu m/s`). Forced unfolding of :math:`Fb` ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ .. raw:: html Forced unfolding of :math:`(Fb)_2` ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ .. raw:: html