Local motions in protein structures
EMSL Project ID
3436a
Abstract
For some years we have been involved in the question of a model for charge transport in peptides. We obtained unusual results for which we have proposed a new theory in 1995 which assigns special local ionization potentials to the amino acids. This in itself is incomplete and our model now postulates that in addition a rapid motion in the
Ramachandran angles. We now proposed that only a very special configuration of protein
segments permits charge transfer. We discovered that a 100 fs time scale for charge
motion results from our MD calculations. The efficiency of charge transport in the
isolated molecule was observed by us experimentally to be very high, however, in water it is known that the efficiency is very low. We recently started large scale MD
calculations which now demonstrated that the loss of efficiency in water can be seen to be
due to a new interference of protein motion as a result of collisions with the hydrophobic
barrel. To obtain these MD results we undertook an important modification of the code of
the CHARMM program to allow for single site excitation, thus defining the time zero
required for the charge to migrate across the peptide. These calculations for water then
resulted in a loss of efficiency of some two orders of magnitude in the MD results. This
corresponded exactly to the known experimental inefficiency of charge transport in water.
Our model thus proposes that there is a special charge transport in proteins due to a
motional induced coupling with a new fast time scale of some 100 fs that is involved in
the background of all protein motions, including protein folding which may have
profound effects on configurational starting points for later folding. Here the influence
of water is calculated on an atomic level, rather than as a global Ansatz, a procedure
which is very intensive in computer time and architecture. However, one of the important problems one must first address in general is whether such
MD studies can be carried out for the isolated molecule, as is typically done, or whether
the presence of water is absolutely required to obtain meaningful results. To investigate
this problem we took a prototype protein as a simple beta hairpin configuration of a
12mer cut from a typical protein and studied the problem in the gas phase. Here we found
some 6 hydrogen bonds in the structure of the hairpin. We then added some 500 water
molecules to the calculation which at once resulted in extending the time for the
computation to some six weeks of computer running time. In fact a 17mer in water
required some 9 months on a single machine. Now it is necessary to not only proceed to
do a particular structure but also to investigate single site variations, analogous to
mutagenesis experiments. Since hydrogen bonds are one of the essential ingredients in the
understanding of protein structure and protein motion such as folding, this demonstrates
that the water medium cannot be ignored, in spite of the enormous cost in computational
time. We need to confirm this finding since it is of substantial import to know if we can
afford to approximate protein motion by neglecting water, in order to make the computation tractable. Unfortunately the results without water are not even approximately
correct ---thus necessitating large scale computations to obtain meaningful results. It is
interesting to note that water helps folding motions, but hinders charge transport.
As helpful as these results are, they point to the definitive need to include water
in any meaningful calculations. These conclusions mean that as a consequence only
supercomputers can provide such answers even for meaningful protein sections. It is
experimentally known that sections of at least 17 amino acids are required to even
approximately model the folding problems encountered in misfolded amyloid systems.
The problem of protein folding is of considerable importance not only for scientific
purposes , but also because over 20 - some people say 100 different diseases are linked to
a false folding of proteins, such as prion diseases leading to the scrapie form, or even variants of Alzheimer or Huntington's disease. To this day it is still a complete mystery
as to the origin of these false folds. Here we have started with a prion as a section of
a 17mer involving the essential proline section. We have abandoned CHARMM as being
too slow and have instead successfully programmed this problem in NWCHEM and as
well as in EGO and NAMD, both of which we found to be some 5 times faster. In
particular we want to also do some local DFT corrections, for which we want to use
CPMD, which is nicely implemented in NWCHEM and EGO, enabling us to combine the
methods. Some preliminary calculations for our amyloid structures reveal that the first 100 nanoseconds appear to be of interest - but here again we find that a single PC can
result in only ps per day or some 8 months for a single calculation in water. Clearly to
attack these problems will require a very large parallel machine which is implemented to
run NWCHEM by the computer support group. This is our present direction and the
motivation of this request
The specific goal of our project is to investigate the anticipated speedup of these
methods. This should involve a checkup both for NWCHEM and EGO, these two being
important for the DFT corrections that can be imbedded to our knowledge only here.
Should these speedups be confirmed and a working methodology established, then we can
start a large scale investigation of the 17mer investigated experimentally in the prion
problem, and in particular to see if the experimental observations on substituent effects
and single site mutagenesis can be confirmed by computations, which would go a long
way towards predicting environmental effects on the prion misfolding problem.
Considering the several thousand papers published each year on such problems, including
Alzheimer's disease, it would be highly interesting to suggest some origins for these
misfolding properties.
Project Details
Project type
Exploratory Research
Start Date
2004-05-04
End Date
2006-06-01
Status
Closed
Released Data Link
Team
Principal Investigator
Team Members