Background and Approach: Protein structure determination by NMR spectroscopy is a time-consuming process which depends upon access to high-field (500-800 MHz) NMR instrumentation for data acquisition, and has historically relied upon on a specialized combination of commercial software and Unix workstations for the subsequent analysis. Investigators who are not equipped with their own NMR instrumentation can take advantage of the National Magnetic Resonance Facility at Madison (NMRFAM), an NIH-funded Biomedical Technology Research Resource Center located in the Biochemistry Department of the University of Wisconsin-Madison. This facility provides a means to acquire the data required for projects involving protein structure determination. However, laboratories which do not routinely engage in structural studies may lack not only the NMR instrumentation, but also the computational tools traditionally used for these projects. Procuring the hardware and software needed for complete NMR analyses may cost more than $10,000 per workstation, a prohibitive expense for some laboratories.
Our goal is to demonstrate a viable alternative which takes advantage of a low cost Unix platform and software tools which are available to academic users at little or no cost.
This approach allows one to use various software packages for biomolecular NMR under the freely available Linux operating system on low-cost Intel-based computers. Software adapted to this approach to date includes NMR processing tools from the NIH (Frank Delaglio), programs developed by groups led by Kurt Wuthrich and Tom Ferrin for spectral analysis and structure determination and software for the 'model-free' analysis of dynamics from relaxation data from Art Palmer. Once the data have been acquired, every step of the structure determination process can be executed in the Linux environment, with no requirement for high-cost workstation hardware.
Our initial development platforms included a single-processor Dell Pentium pro 200 machine and an Intel 4 x Pentium pro 200 workstation running the RedHat 5.2 distribution of Linux. Over the past 7 months we have performed an extensive 'real-world' demonstration of this concept using a Dell Inspiron 3500 laptop computer for completely portable structure determination with a collaborator at the University of Rochester. We acquired a second laptop in July 1999, a Dell Inspiron 7000, and have updated most of these machines to RedHat 6.0 or 6.1. Our implementation and descriptions are geared toward the users of the Bruker DMX spectrometers at the National Magnetic Resonance Facility at Madison (NMRFAM), but should be generally applicable to anyone engaged in structural biological NMR research.
Protein Structure Determination in a Nutshell: The basic steps in determining an NMR structure include sample preparation and data acquisition (which occur before you encounter this set of tools), processing the raw (time-domain) data by FT into frequency-domain spectra, peak-picking, resonance assignment, NOE restraint generation and structure calculation. The final two steps are generally performed in an iterative manner, with the result of one round of structure calculations used to help correct misassigned NOE crosspeaks and identify new NOE restraints for the next round of structure calculations. In the LinuxNMR project we have successfully executed all of these stages of a new protein structure determination using non-commercial software on a laptop computer which cost well under $3000.
The linuxnmr@nmrfam.wisc.edu mailing list is a general forum to discuss the usage and development of applications for biomolecular NMR spectroscopy in the Linux environment. Users are encouraged to post questions and comments so that solutions to common problems may be disseminated and stored in a freely accessible archive.
Posting to the LinuxNMR mailing listIf you send mail to linuxnmr@nmrfam.wisc.edu, your message will be distributed to all subscribers on the list.
Web access to the LinuxNMR mailing listThis mailing list is archived and may be accessed via the web at:
http://www.nmrfam.wisc.edu/Listarchives/linuxnmr-archive/maillist.html
Subscribing to the LinuxNMR mailing listIf you want to add yourself to the linuxnmr mailing list, send a message with the word subscribe in the subject or body of the message to:linuxnmr-request@nmrfam.wisc.edu:
Unsubscribing from the LinuxNMR mailing listTo unsubscribe from the linuxnmr mailing list, send a message with the word unsubscribe in the subject or body of the message to:
linuxnmr-request@nmrfam.wisc.edu
Use the linuxnmr-request address and NOT the list address (linuxnmr) for all subscribe/unsubscribe requests.
This section provides an inexhaustive but representative chronology of LinuxNMR development history at NMRFAM: site updates, new projects, personnel, etc.2/16/00 - Well, the paper's not entirely done, but I couldn't resist adding an image of the new Dcp structure to the front page. Stay tuned for more information on the research projects supported by LinuxNMR.
1/12/00 - Another LinuxNMR protein structure is completed, the 81 residue protein mentioned earlier, by Brian. Illustrated descriptions of LinuxNMR-related research projects will be added once the paper is out of the way.
11/11/99 - Irene Ong, a CS graduate student working with NMRFAM, has successfully implemented DYANA in the CONDOR environment for high-throughput computing on workstation clusters. Hopefully we will soon have access to a Linux CONDOR cluster which will provide a platform for highly parallelized DYANA calculations.
11/4/99 - The first protein structure determined with LinuxNMR is deposited in the RCSB protein databank by Betsy Lytle - a dockerin domain from the cellulosome of the thermophilic bacterium C. thermocellum1dav.pdb (ensemble) and 1daq.pdb (minimized average)
10/1/99 - After a couple of weeks in Europe, Brian returns to begin work on the structure of an 81 amino acid protein, in collaboration with a group from Northwestern University, using the Inspiron 7000 for nearly all NMR processing, analysis and structure calculation tasks. A portion of the 3D data processing was performed earlier in the project using Felix 95.0 on the SGI. All analysis of 3D 15N and 13C NOESY spectra is performed in XEASY, and structures will be calculated with DYANA.
8/23/99 - Initialized LinuxNMR mailing list (linuxnmr@nmrfam.wisc.edu) for recording useful information on software development and implementation by users.
8/3/99 - Used Dell Inspiron 7000 laptop in LinuxNMR demonstration for NMRFAM NIH site visit review (with assistance from Sonay Kuloglu), showing XEASY analysis of 2D and 3D NMR data as well as structural analysis with MOLMOL.
8/1/99 - Installed ModelFree 4.01, software for analysis of 15N relaxation data from Prof. Arthur Palmer, on the Inspiron 7000, to begin model-free analysis of data on the dockerin domain and other proteins.
7/13/99 - Received 2nd LinuxNMR portable: a Dell Inspiron 7000, 366 MHz Intel Celeron with 192 MB of memory and a 10 GB HD.
7/1/99 - Structure refinement continues apace in Rochester on tuni.
7/1/99 - Thomas Goddard (UCSF Computer Graphics Lab) delivers a prerelease update of Sparky, his program for analysis of NMR data, which incorporates all aspects of 15N relaxation analysis (peak picking and exponential curve-fitting). This is a common task which we have previously performed with commercial software, but can now be moved to this academic software tool and added to the LinuxNMR suite.
6/20/99 - Ordered RedHat 6.0 CDs to upgrade LinuxNMR machines.
5/18/99 - Ordered 2nd LinuxNMR portable: a Dell Inspiron 7000, 366 MHz Intel Celeron with 192 MB of memory and a 10 GB HD.
5/11/99 - Betsy Lytle returns to Rochester, NY with tuni to continue structure refinement. Tuni is on the net via cable modem with RoadRunner (TimeWarner) by 9:30 a.m.
5/3/99 - Betsy Lytle comes to NMRFAM in Madison, WI to begin refining the dockerin structure on tuni.
4/27-5/3/99- Sonay Kuloglu (Markley lab grad student) beta tests tuni for assignment of 3D triple-resonance spectra of her protein using XEASY.
4/12-4/26/99- Brian alpha tests tuni for all aspects of NMR data processing, analysis and structure calculation.
4/8/99 - Steve has installed/successfully used these software tools on tuni:
- nmrPipe
- xeasy
- dyana
- molmol
- spscan
3/25/99 - Took delivery of 1st Dell Inspiron 3500: Tuni is a 300 MHz Intel Pentium II processor laptop with 96 MB of memory and a 4 GB HD.
3/25/99 - Got SPSCAN working on doran for conversion of processed nmrPipe data to XEASY submatrix format. /home/volkman/data/nmrpipe/pipe2xeasy_a.spm successfully converts a 2D NOESY spectrum to Xeasy format with referencing intact. This file (test.ft2) was converted (bruk2pipe) and processed (nmrPipe) on doran (Linux). Spscan needs to be linked into the main /usr/local/bin on doran.
2/8/99 - Betsy Lytle, a University of Rochester graduate student in the lab of Prof. J.H. David Wu, comes to NMRFAM to begin work on the assignment and structure determination of the dockerin domain from C. thermocellum, in collaboration with Milo Westler and Brian Volkman. By the end of her visit Betsy has been sufficiently trained in NMR assignment methods that she could continue work on the project in Rochester, however the necessary computers and software are not available in her lab or department. We begin searching for the most efficient method to export this infrastructure to allow her to finish the project back home. Because we have chosen to use academic software for the main tasks of the project (XEASY and DYANA for assignment and structure calculations), and we have already obtained (or compiled) Linux versions of these programs, the concept of a laptop PC running Linux is suggested as a low-cost portable alternative to the purchase or loan of a full-size desktop Unix workstation.
Brian Volkman is an asst. scientist at NMRFAM with experience in multidimensional NMR spectroscopy and protein structure determination. He and Milo Westler (NMRFAM Operations Director) conceived the LinuxNMR project while brainstorming for a way to provide an external collaborator with a computational platform that could bring all the tools necessary for a protein NMR structure determination project together in a easily transportable workstation.Steve Wilkens is a biophysics graduate student in the lab of Prof. John Markley. In his spare time, he does density functional calculations on the active sites of iron-sulfur proteins in order to determine NMR observables like Fermi contact shifts and relaxation times. Steve puts the Linux in LinuxNMR.