FOR THE PRE-FINAL IN-CLASS ORAL REPORT:

Basically, you need to show you are on track. 
Show at least one computtational protocol, 
the trajectory (energy(time) plot), and 
your lowest energy structure. 3-5 slides. 

This would be a good time to ask questions and 
spell out problems you might have encountered to
get feedback from class and your instructor. 



FOR THE FINAL WRITTEN REPORT: 

1. Summarize you mid-term findings. 

2. Explore 2-3 different protocols of simulated annealing folding. 
Here is one idea. Run a fairly long MD simulation at the protein
melting temperature (317K), collect several (10) lowest energy structures, 
then cool each of them to 300K, run at 300K for a while, 
and then get a few lowest energy ones 
from those trajectories. Compare them visually, see if they make sense 
based on your lattice simulation results.

3. Run a long simulation at the melting temperature, output energy vs. time. 
Compute radius_of_gyration(time). Plot them together. Do you see
folding/unfolding events? Confirm by visualizing the structures via VMD. 
Use a representation that shows secondary structure. You can use the 
radius of gyration .awk script from the lectures. 

4. Make sure your report contains pictures and energy plots + careful 
description of the annealing protocols, use graphics to schematically
represent them. 


Technical notes: 

a. See "Protein folding. Unfolded starting structures + some useful MD
protocols." link on the class site. You will find specific annealing input 
file examples there. MAKE SURE TO USE THE CORRECT ENVIRONMENT VARIABLES AND 
CUDA_VISIBLE_DEVICES from the working example in /EXAMPLES 
(See also your homework and AMBER tutorial you have completed.) 

You can modify the temperature values as well as 
the duration of each annealing phase (value1, value2 and istep parameters). 
I do not recommend changing 
anything esle unless you know what you are doing, except, maybe,
the parameter gamma_ln. Lower values of gamma_ln give you lower viscocity, 
that is faster folding. You can use gamma_ln=0.01. Remember that 
lower gamma_ln also means weaker copupling to your thermostat, the 
characteristic time [in picoseconds] is 1/gamma_ln. Thus, make sure that if 
gamma_ln is very small, the system still 
reaches the desired temperature over the 
specified annealing time.  

b. To take a given point from the AMBER trajectory and make it a starting
point of a new simulation, you can do the following. Once the original 
trajectory is generated, input it into the ptraj script from 
your example (in the home directory). Modify the last line to "trajout a.rst
restart", which will force the script to output a.rst.1, a.rst.2, etc. files
instead of PDB files. Select frame number "N" you want, use it as 
your sander input in "-c a.rst.N"   (See e.g. cuda_annealing.sh. 
Check if you need to remove "N" in "a.rst.N" ).   

c. You do not have to run everything on GPU using $AMBER/bin/pmemd.cuda
   and wait in line till your card frees up. You can test everything
   using $AMBER/bin/sander, which will run on a single CPU instead (you do 
   not need to reserve time for it). Just make sure you set nstlim 
   variable to some small number of steps, say nstlim=100 , so your 
   job does not take too long. Make sure to adjust your print frequency
   accordigly, e.g. ntpr=100, ntwx=100. You can test many features 
   of your protocol this way, e.g. figure out the memory cost per 
   100 structures (ntwx=100) saved, etc. You won't be able to fold
   your protein though, for that you need the "supecomputer" mode.