Computational Algorithms to Characterize Biomolecular Mechanisms and Allostery
How do biomolecules change shape and perform function? How can we identify collective motions and excited substates from sparse, spatiotemporally averaged data?
Our lab develops fast algorithms to structurally characterize biomolecular ensembles. Methods and algorithms from robotics and machine learning have proved powerful in modeling protein structure, collective motions, and uncovering allosteric sites. We create and apply these algorithms and procedures to probe conformational distributions, to interpret experimental (crystallography and NMR) data, and identify molecular mechanisms.
Collision free Poisson motion planning in ultra high dimensions
J Comput Chem 10.1002/jcc.25138 (2018) Biomolecules are nano-machines. And, like self-driving cars, they need to navigate a complex, high-dimensional landscape to perform their function. Navigating molecular conformational spaces with randomized algorithms can provide a fast alternative to more expensive molecular dynamics simulations, especially for insight into large-scale motions. However, such algorithms often don't scale well with the ultra-high dimensionality of configuration spaces (1,000s of degrees-of-freedom!), and self-collisions lead to low acceptance rates of trial moves. We present two novel mechanisms to overcome these limitations. First, we introduce a new move set by adding temporary constraints between near-colliding atoms in a trial move. The resulting co-dimension one constraint manifolds instantaneously redirect the search for collision-free conformations, and couple motions between distant parts of the linkage. Second, we adapt a randomized Poisson-disk motion planner, which prevents local oversampling and widens the search, to ultra-high dimensions. Tests on several model systems show that the sampling acceptance rate can increase from 16% to 70%, and that the conformational coverage in loop modeling measured as average closeness to existing loop conformations doubled.
KGSrna Characterizes Excited State of HIV-1 TAR
Nucl. Acids Res. 42:9562-9572 (2014) KGSrna efficiently probes the native ensemble of RNA molecules in solution. KGSrna ensembles accurately represent the conformational landscapes of three-dimensional RNA encoded by NMR proton chemical shifts. KGSrna resolves motionally averaged NMR data into structural contributions; when coupled with residual dipolar coupling data, a KGSrna ensemble revealed a previously uncharacterized transient excited state of the HIV-1 trans-activation response (TAR) element stem-loop.
CONTACT: Collective Motion from Multiconformer Models
Nat Meth 10:896-902 (2013) CONTACT analyzes protein structures determined by room temperature X-ray crystallography. Built upon our earlier computational procedure qFit, CONTACT detects how subtle features in the experimental data produced by the conformational ensemble propagate through the protein and identifies regions within the protein where these cascades of small changes are likely to result in stable conformations. Read more ...