LFBGS

[alberto99]

Hi Justin,
When we start simulations from a sequence in which there is a PRO followed by a TYR there's often an overlap of atoms when starting from LEAP.
The normal minimization scheme we have does not get rid of this.
Amber's L-BFGS does. I see that there is the option:

LocalEnergyMinimizer.minimize(context)

And i've seen that you and John have tried it and think is slow. I wonder if it is worth to have an option in MELD to add it.

Here is a sequence that triggers this behavior:
AARYDDVLYFPASRYPETGAHISDAIKAGHADVCTIERSGADKRRQESLKGIPTKPGFDRDEWPMAMCEEGGKGASVRYVSSSDNRGAGSWVGNRLNGYADGTRILFIVQ
Al.

[jlmaccal]

What is the normal minimization scheme we do? I can’t remember if we use Amber of openmm. I’m confused by what you mean about the code snippet.

Would it be possible to find a chi1 angle that would get rid of the clash? Then we could just tweak that angle prior to minimization. This would be my preferred solution.

On Jul 2, 2014, at 3:29 PM, alberto99 notifications@github.com wrote:

Hi Justin,
When we start simulations from a sequence in which there is a PRO followed by a TYR there's often an overlap of atoms when starting from LEAP.
The normal minimization scheme we have does not get rid of this.
Amber's L-BFGS does. I see that there is the option:

LocalEnergyMinimizer.minimize(context)

And i've seen that you and John have tried it and think is slow. I wonder if it is worth to have an option in MELD to add it.

Here is a sequence that triggers this behavior:
AARYDDVLYFPASRYPETGAHISDAIKAGHADVCTIERSGADKRRQESLKGIPTKPGFDRDEWPMAMCEEGGKGASVRYVSSSDNRGAGSWVGNRLNGYADGTRILFIVQ
Al.

—
Reply to this email directly or view it on GitHub.

[alberto99]

we use openmm's
simulation.minimizeEnergy()
which is not L-BFGS.

You are right though, it looks like a chi1 movement would get rid of the clash.

Al.

Alberto Perez (Homepage)
Laufer Junior Fellow
Laufer Center for Physical and Quantitative Biology
Stony Brook University
Watch my TEDx talk: You can't always blame your parents!

On Jul 2, 2014, at 5:46 PM, Justin MacCallum notifications@github.com wrote:

What is the normal minimization scheme we do? I can’t remember if we use Amber of openmm. I’m confused by what you mean about the code snippet.

Would it be possible to find a chi1 angle that would get rid of the clash? Then we could just tweak that angle prior to minimization. This would be my preferred solution.

On Jul 2, 2014, at 3:29 PM, alberto99 notifications@github.com wrote:

Hi Justin,
When we start simulations from a sequence in which there is a PRO followed by a TYR there's often an overlap of atoms when starting from LEAP.
The normal minimization scheme we have does not get rid of this.
Amber's L-BFGS does. I see that there is the option:

LocalEnergyMinimizer.minimize(context)

And i've seen that you and John have tried it and think is slow. I wonder if it is worth to have an option in MELD to add it.

Here is a sequence that triggers this behavior:
AARYDDVLYFPASRYPETGAHISDAIKAGHADVCTIERSGADKRRQESLKGIPTKPGFDRDEWPMAMCEEGGKGASVRYVSSSDNRGAGSWVGNRLNGYADGTRILFIVQ
Al.

—
Reply to this email directly or view it on GitHub.

—
Reply to this email directly or view it on GitHub.