274. On the Spectral Projected Gradient Method for the Molecular Distance Geometry Problem
Invited abstract in session TB-13: Numerical Methods and Applications I, stream Numerical Methods and Applications.
Tuesday, 10:30-12:30Room: B100/6009
Authors (first author is the speaker)
| 1. | Mariana da Rosa
|
| Institute of Mathematics, Statistics and Scientific Computing, University of Campinas (UNICAMP) |
Abstract
The reconstruction of 3D protein structures from Nuclear Magnetic Resonance (NMR) data presents significant challenges due to uncertainties in the distance constraints between atoms within a protein molecule. We use a Spectral Projected Gradient algorithm to address these challenges, overcoming the limitations of previous methods that unrealistically assume all pairwise interatomic distances are known, are restricted to small problem instances, or only consider narrow distance ranges. Our approach incorporates fixed covalent bond lengths and angles while effectively handling larger interval uncertainties, offering a more accurate representation of the experimental conditions in NMR data. Computational experiments highlight the potential of this method to provide reliable solutions for molecular structure determination.
Keywords
- Applications of continuous optimization
- Computational mathematical optimization
- Global optimization
Status: accepted
Back to the list of papers