Each of the structures generated by the NMR study was used as the starting 
structure for an independent simulation.  The N-acetyl galactosylated peptides
were each run twice, with different random number sets to ensure they sampled
different conformational spaces.  Thus, there were 2x16 GalNAc runs, 36 Man
runs and 48 for the unglycosylated peptide, for a total of 116.  Each initial
structure was treated in the manner described below.  Some runs were not 
completed when the data presented here were extracted for analysis (27 
December, 2012).

AMBER prmtop and inpcrd input files were generated using the tleap utility in 
AmberTools12.  Each PDB file provided by the NMR software was used as initial 
input.  To it were added sufficient chloride counter ions to ensure neutrality.
The neutralized system was solvated using small, pre-equilibrated boxes of 
TIP3P water to form a solvent buffer of at least 10 Angstroms from each atom 
in the system.  To ensure free movement, the box was increased so that it was 
approximately cubic, using the longest initial dimension as the target for all 
dimensions.  The peptide portions of the molecule were modeled using the 
ff12SB force field, the ions were modeled using the ions08 parameters, and the 
monosaccharides were modeled using GLYCAM_06h-1 with GLYCAM_06h-12SB amino 
acid atom types.  The complete set of parameters and setup files are available 
upon request from author Foley.

Each simulation was begun with a steepest-descent minimization for 10,000 steps.
The time step for all subsequent phases was 0.002 ps, and bonds involving 
hydrogen were constrained according to the SHAKE algorithm.  The systems were 
run in the NTP ensemble using Particle Mesh Ewald to treat electrostatics.  
Next, the system was heated gently in 20,000 steps (40 ps) from 5 to 300 K at 
1 atm using the weak-coupling algorithm.  In the equilibration and production 
phases, the temperature was maintained at 300 K using  Langevin dynamics.  
After heating, the systems were allowed to equilibrate for 2 ns before 
beginning a 100 ns production phase.  Due to simulation restarts, a few systems 
ran slightly longer.  System coordinates were saved every 5000 steps (10 ps)
during the production phase.  The results presented here are from the final, 
production phases, and unless otherwise noted represent the collected data 
from each set of runs.