TY - JOUR
T1 - Two-dimensional grid optimization for sedimentation velocity analysis in the analytical ultracentrifuge
AU - Kim, Haram
AU - Brookes, Emre
AU - Cao, Weiming
AU - Demeler, Borries
N1 - Publisher Copyright:
© 2018, European Biophysical Societies' Association.
PY - 2018/10/1
Y1 - 2018/10/1
N2 - Sedimentation velocity experiments performed in the analytical ultracentrifuge are modeled using finite-element solutions of the Lamm equation. During modeling, three fundamental parameters are optimized: the sedimentation coefficients, the diffusion coefficients, and the partial concentrations of all solutes present in a mixture. A general modeling approach consists of fitting the partial concentrations of solutes defined in a two-dimensional grid of sedimentation and diffusion coefficient combinations that cover the range of possible solutes for a given mixture. An increasing number of grid points increase the resolution of the model produced by the subsequent analysis, with denser grids giving rise to a very large system of equations. Here, we evaluate the efficiency and resolution of several regular grids and show that traditionally defined grids tend to provide inadequate coverage in one region of the grid, while at the same time being computationally wasteful in other sections of the grid. We describe a rapid and systematic approach for generating efficient two-dimensional analysis grids that balance optimal information content and model resolution for a given signal-to-noise ratio with improved calculation efficiency. These findings are general and apply to one- and two-dimensional grids, although they no longer represent regular grids. We provide a recipe for an improved grid-point spacing in both directions which eliminates unnecessary points, while at the same time providing a more uniform resolution that can be scaled based on the stochastic noise in the experimental data.
AB - Sedimentation velocity experiments performed in the analytical ultracentrifuge are modeled using finite-element solutions of the Lamm equation. During modeling, three fundamental parameters are optimized: the sedimentation coefficients, the diffusion coefficients, and the partial concentrations of all solutes present in a mixture. A general modeling approach consists of fitting the partial concentrations of solutes defined in a two-dimensional grid of sedimentation and diffusion coefficient combinations that cover the range of possible solutes for a given mixture. An increasing number of grid points increase the resolution of the model produced by the subsequent analysis, with denser grids giving rise to a very large system of equations. Here, we evaluate the efficiency and resolution of several regular grids and show that traditionally defined grids tend to provide inadequate coverage in one region of the grid, while at the same time being computationally wasteful in other sections of the grid. We describe a rapid and systematic approach for generating efficient two-dimensional analysis grids that balance optimal information content and model resolution for a given signal-to-noise ratio with improved calculation efficiency. These findings are general and apply to one- and two-dimensional grids, although they no longer represent regular grids. We provide a recipe for an improved grid-point spacing in both directions which eliminates unnecessary points, while at the same time providing a more uniform resolution that can be scaled based on the stochastic noise in the experimental data.
KW - Analytical ultracentrifugation
KW - Computational biophysics
KW - High-performance computing
KW - Hydrodynamics
KW - Numerical optimization
UR - http://www.scopus.com/inward/record.url?scp=85047123646&partnerID=8YFLogxK
U2 - 10.1007/s00249-018-1309-z
DO - 10.1007/s00249-018-1309-z
M3 - Article
C2 - 29777290
AN - SCOPUS:85047123646
SN - 0175-7571
VL - 47
SP - 837
EP - 844
JO - European Biophysics Journal
JF - European Biophysics Journal
IS - 7
ER -