I anticipate lacking of prior experience with dimensionality reduction problems.
Some scientists concerned with drug discovery performed several steered
Molecular Dynamics simulations of the
alanine-dipeptide molecule dragged by a radial force from an equilibrium
conformation to another different equilibrium conformation.
They sampled at regular intervals 7 dihedral angles, 5 bending angles, and 4
atom pair distances all along the trajectory from
the initial to the final conformation. Likewise they also calculated the work
done on the molecule by the applied force.
When a real system is simulated (this is a toy model) macro-molecules
conformational space is huge. Not all the sampled variables
change coherently during the chemical-physical reaction (conformation change)
taking place.
A dimensionality reduction is necessary.
We need to identify the subset of variables that are necessary and sufficient
to describe such a reaction.
Such variables must exhibit the highest correlation with the work curves.
The objective is to correlate the work curves with specific structural
descriptors (variables like angles, distances, and so on)
and automatically discriminate between descriptors that correlate with work
curves and descriptors which
are not involved in the reactive pathway.
I thought I could perform a dimensionality reduction first (experimenting with
some common non-linear methods) and then perform regressionto estimate the
correlation (dependence) of the survived variables with the work curves.
Then i came across R package "dr" which apparently does it all in one step.
Unluckily the examples that come with dr documentation are far away from the
research field I am involved in.
I know dimensionality reduction methods apply to many different fields.
I wonder whether dr can help me with finding a reduced representation of
molecule conformational changes.
Data samples are available upon request.
Thank you in advance.
Maura
tutti i telefonini TIM!
[[alternative HTML version deleted]]
______________________________________________
[email protected] mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
