Hybrid refinement of heterogeneous conformational ensembles using spectroscopic data Conformational ensemble estimate Jennifer M. Hays University of Virginia Blue Waters Symposium 2019 x 2 x 1 x N
Proteins exhibit a broad range of flexibility http://www.ibs.fr/research/research-groups/protein-dynamics-and- fl exibility-by-nmr-group-m-blackledge/
Estimating the Duhovny, Kim, & Sali, BMC Structural Biology, 2012 conformational ensembles of flexible proteins: a difficult inverse problem Experimental data tend to come in two Single varieties: structure prediction 1. Ensemble average quantities (NMR, from SAXS SAXS). 2. Distributional data that are sparse over the atomic coordinates (DEER, FRET). Sparse labels from DEER Jeschke, Protein Science, 2017
Estimating the conformational ensembles of flexible proteins: a difficult inverse problem Experimental data tend to come in two varieties: Lot’s of great work has been done to leverage 1. Ensemble average quantities (NMR, ensemble average SAXS). quantities 2. Distributional data that are sparse over These data are harder to the atomic coordinates (DEER, FRET). deal with!
Bias-resampling ensemble refinement (BRER) Conformational 1 2 ensemble estimate ... x 2 (1) Draw N conformations N x 1 from with replacement x N (2) Sample N distanc- es from the experi- mental distribution. Refine each confor- mation X n against one Update conformational distance d n estimate with resulting ensemble d 2 d 1 d N P DEER (d) Compare estimated distributions to experiment Hays, Ca fi so, & Kasson, JPC Letters, 2019
Bias-resampling ensemble refinement (BRER) A) Iteration 1: 5 Resample a Conformational Iteration 2: Iteration 100: 1 2 targets subset of ensemble estimate 10 targets 500 targets weighted by (aggregate) (aggregate) ... x 2 Probability (1) Draw N conformations N x 1 from with replacement x N (2) Sample N distanc- es from the experi- mental distribution. Refine each confor- Distance mation X n against one Update conformational distance d n B) estimate with resulting ensemble d 2 Target Distance (nm) d 1 Training d N Convergence Production d MD U bias = α d target P DEER (d) Compare estimated distributions to Time (ns) experiment Hays, Ca fi so, & Kasson, JPC Letters, 2019
Syntaxin and SNAREs SNARE
Syntaxin and SNAREs Dawidowski and Ca fi so, Biophys J., 2013
ained−ensem A) BRER EBMetaD restr ained−ensem ble 1.00 BRER refinement of 52/210 0.75 syntaxin 0.50 conformational 0.25 ensemble reproduces 0.00 1.00 the experimental 105/216 Probability 0.75 distributions very 0.50 well 0.25 Distance (nm) B) 0.00 0.100 1.00 Divergence *** p < 0.001 0.075 196/228 0.75 0.050 *** 0.025 *** *** 0.50 0.000 0.25 052/210 105/216 196/228 0.00 Dawidowski and Ca fi so, Biophys J., 2013 0 2 4 6 0 2 4 6 0 2 4 6 Hays, Ca fi so, & Kasson, JPC Letters, 2019 Distance (nm) B) ergence
Open state A) B) Partially Linker open state 196 196 BRER refinement 52 210 216 reveals 228 216 Habc previously H3 228 105 unresolved C) 0.8 52/210 105/216 196/228 partially open 0.6 Probability syntaxin states 0.4 0.2 0.0 0 2 4 6 0 2 4 6 0 2 4 6 Distance(nm) Hays, Ca fi so, & Kasson, JPC Letters, 2019
Thank you! • Kasson Lab • Cafiso Lab • Dave Cafiso • Peter Kasson • Damian Dawidowski • Eric Irrgang • Ania Pabis • Anjali Sengar • Ricardo Ferriera
Thank you! • Kasson Lab • Cafiso Lab I use Blue Waters because… • Dave Cafiso • Peter Kasson - 15 μs of all-atom simulation data for this project • Damian Dawidowski - Over the course of my fellowship used over 200k • Eric Irrgang node-hours Hays et al., Ang. Chemie, 2018 Hays, Ca fi so, & Kasson, JPC Letters, 2019 - Support team • Ania Pabis • Anjali Sengar • Ricardo Ferriera
Recommend
More recommend
