Folding a miniprotein, a starter protocol
===================================

Input files
-----------
- minip.top : topology file of protein T (Amber format. to be used with the standard energy function, igb=5 )
- minip.altE.top  topology file of protein T (Amber format. To be used with
  an alternative, more advanced energy function, igb=8. 
Note that all of the examples below assume igb=8 )
- minip.crd : coordinates of atoms, extended, unfolded structure (Amber
  format. For the alternative energy function this file is the same, but I
created a copy of it with "AltE" name for convenience )
- minip.pdb : PDB file plus charges and radii of atoms

OPTIMIZATION STAGES: 


STAGE 0. Run minimization with sander. This is to remove any "hot spots" in 
the above structures.  Always do it. First step if you start with a
"manually" prepared structure, such as the completely extended protein. 


----------------------------
500 steps of steepest descent algorithm
500 steps of conjugate gradient algorithm
./sander_min.sh

---------
Below are bits and pieces of computational protocols, stage by stage, 
that you may find 
useful. You will need to modify them accordingly 
-------------------------------------------------------

STAGE 1. 

Heating from 0k to 317k (experimentally known melting temperature of protein T)
-------------------
30ps of heating (from 0k to 317k)
20ps of md simulation at constant temperature T=317k
./cuda_heat.sh

STAGE 2. Run 20ns of simulation. This is "mild" simulated annealing. 
-----------------------
20ns of MD simulation at 317K (melting temperature, slightly higher than
room temperature.)
It may already be possible to see folding unfolding events at this stage, 
especially if we increase the simulation to 100ns or more. 
I recommend to run for about 1/2 day to collect several folding events. 

./cuda_equil.sh

STAGE 3. (Strong) Simulated Annealing
-------------------
Increase the temperature from 317k to 400k , 3ns
Then decrease temperature from 400k to 317k, 4ns
MD simulations at 317k , 2ns
./cuda_annealing.sh

STAGE 4. Run 10ns of simulation. This is "mild" simulated annealing.
--------------
10ns of Molecular dynamics at 317K (melting temperature, slightly higher than
room temperature.)
You may want to do a longer simulation (100ns)
./cuda_prod.sh


So, your first possible protocol may look like this: 

STAGE0 -> STAGE1 -> STAGE2 -> STAGE3 -> STAGE4 -> analyze energies (EPtot) from
stage 4, choose the structure with the lowest EPtot energy.
Note that output files from stage n become inputs for stage n+1. Look inside the corresponding shell scripts to see how it works. You may create
a single shell script to run the 4 stages automatically. 

As a second step beyond the basic, 
you may create a loop, where you take the output from
STAGE4 and feed it into the input of STAGE1, repeat K > 1 times, analyze and
compare structures from all K STAGE4 runs you collect. 
  
Then, you can play with the high temperature used in STAGE 3. Perhaps go
lower than 400K? Or, increase the length of STAGE 4?   Also, there is the
"gamma protocol" mentioned in the other document.

Generally, think of the different STAGEs as somewhat maulliable lego blocks.  
