Friday, September 19, 2008

Dr. Kolossavary, DESMOND, bilipid membranes, and other great ways to spend a weekend

I am behind the times so often it is startling. Dr. Kretsinger took some time on a recent hike to up it to me bluntly "It would help if you caught up to physics 1908." With that in the background, attendance at the Chem E. lecture - arranged by the Shirts Group - made for an enjoyable blitz-crash.

One of my favorite facets of America is the influx of great minds generated of great cultures. Kolossvary is of Hungarian descent, and very much looks the part - broad shoulders, thin mustache, and Mediterranean complexion... with one notable exception... the fingers of a programmer. Much like a pianist, a programmers fingers, albeit strong, are often long and dextrose. In any case Dr. Istvan fits the part of the guy who has sledge-hammered on a chemo-physio-computer engineering project around for the past 17 years. My guess is this guy's keyboards last ~3-5 months max.

At first blush DESMOND has some impressive stats. Dr. Shirts of folding@home fame chuckled throughout the lecture. At this slide he cited the first microsecond of modeling required 6 months to calculate.
Towards the bottom of the slide note the Anton - this processor, also developed by D. E. Shaw Research, works in tandem with Desmond to crunch up to 14 microseconds per day.

To facilitate we amateurs Desmond comes with a free downloadable Desmond GUI merged with Schrodinger's Maestro software. An impressive and thorough package a full license for Maestro runs in the $30K+ realm. The academic license, however, is free and provides an excellent platform to enjoy Desmond from. The pros: In fifteen minutes or so I had gone through some basic protein preparation wizard and was running Desmond through operations and getting a bi-lipid membrane like the pictured above. Amazing stuff that might have taken every bit of a month a few years ago to understand and initiate now accomplished in minutes.

Beneath the GUI hood there are even more features. To access these Desmond also comes with a command line interface - this provides additional algorithms for more in depth view and control. Navigating between the gui and command line is helped greatly by Maestro's process and file management that creates files prior to executing actions that are very simple to read and modify. Shirt enthusiastically likened these unto a voting paper trail, and I must agree it's a handy feature.

Cons: well it's not too big a deal, however there's no 64 bit binary yet, and for simplicities sake I ended up running Maestro in a virtual 32 bit machine to get everything more stable. Dr. Kolossvary indicated that this upgrade might be coming as early as this month - so stay tuned. Another planned feature includes direct reading of OPM's protein database files.

This week is finally calming down and am going to start running some simulations - stay tuned for part 2 some time in the next week or so.

The Making of Membranes
  1. After importing a PDB file go to the protein preparation wizard under "workflows" on the far right hand side
  2. Dr. Kosovarry recommended assigning bond orders, adding hydrogens, treating metals, capping the termini, and detecting the disulfide bonds. In his example he did not feel it necessary to delete waters... there are some other options below that are only available with the full version.... ah well... it's a start. Also there appear to be some missing AA groups and I am yet to figure out how to fix that...
  3. Now under "Applications" go to "Desmond" and go to the System Builder. There are 3 solvent models available, with more on the command line - There's an ion placement submenu perhaps a ligand placement menu may appear in future versions.
  4. The membrane builder menu let's you autoposition the membrane using selected atoms or amino acids - I've had moderate success in this using the selection tool to grab a bunch of atoms around the disc created by the OPM database, If you're not satisfied with the initial placement modify it clicking the "Adjust..." checkbox.
  5. When you're ready to see your model click "Start" - depending on a 2cpu 3.2 ghz machine you'll see a model like the above in a minute or two with ~120k atoms in the case of FAAH.

No comments: