Fast NMR Methods at ICMRBS 2014

On Wednesday the 27th August, we wil be presenting a workshop on Fast NMR Data acquisition methods, with a focus on GFT data acquisition, nonuniform sampling and subsequent processing. Below is a brief outline of what we will be presenting. There will be an opportunity to work on your own data so bring along your Linux or Mac OS X laptop with nmrPipe already installed. See below for details on getting the hmsIST reconstruction software that works along with nmrPipe. However, a laptop with software is not neccessary to participate.

Background:

We will be briefly introduce some of the different fast NMR methods used today for biomolecular structure determination followed by history, theory and context for GFT and nonuniform sampling (NUS) techniques. What are the various ways to quicken the pace of data collection? What are the implications for various methods? If we are going to skip points, what points are the best points to acquire? How fast can we go? These are the things we will discuss.

NUS Pulse Sequences and Utilizing Poisson Gap Sampling Schedules:

We will present a quick tutorial on how to convert Topspin 2.1 pulse sequences into NUS sequences and how to implement Poisson Gap Sampling Schedules with them.

We will continue with the setting up of NUS experiments with Topspin 3.0+ and Agilent sotfware so they can use Poisson Gap Sampling Schedules.

How to Process NUS Data:

A large amount of the workshop will cover processing of Nonuniform Sampled data. That is, how do we recover a useable spectrum from the data we did acquire?

We will primarily discuss processing with hmsIST and nmrPipe. Examples will include 2D and 3D spectrum processing, both of which can be performed easily on modest laptop computers running Linux or Mac OS X. Bring your laptop!!! Common mistakes will be covered and knowledge on how to process data with different acquisition modes will be developed.

BYO Linear Data!:

Bring along a regular data set. We will show you how to create a NUS version of the data and then you can attempt to reconstruct it. Of course if you have been acquiring nonuniform data and have been having trouble with reconstruction, bring it along and we will get you up and running.

Get The Software:

For the hmsIST reconstruciton software, contact . If you would like the Topspin 3.0+ Macro that automatically sets up Poisson Gap Sampling schedules, contact for the package.

Combination of NUS with other fast methods:

NUS can be combined with other fast NMR methods such GFT NMR. This will be briefly discussed at the workshop. For more details, see here.

 

How IST works... A brief Outline rollup

A fourier transform requires complex points with an evenly spaced time interval between each point. In Non Uniform Sampling, we violate this requirement by skipping points. The sparcity of NMR signals allows for us to do this, however we can not simply apply a Fourier transform to the points we have collected. Even if we replace the skipped points with zeros we get a transform which may not look too bad, but does contain artifacts. See below:

 

An example of a clean uniformly sampled signal. All samples are included in the transform.

An example of what happens when you randomly set samples to zero. Here 50% fo samples have been set to zero and a fourier transform performed on the resultant signal.

The signal can be recovered without the resultant baseline "noise" (this is the result of the point spread function of the sampling schedule). The method descrubed here is the Iterative Soft Threshold (IST) method. A threshold cutoff is applied to a fourier transformof the non uniform signal and (see below, left) with data above the threshold placed in a buffer (see below, right). This buffer is inverse fourier transformed and this signal is then subtracted from the non uniform signal as oringally acquired. This new signal is then forward fourier transformed and a new lower threshold applied.

 

Reconstruction of a nonunform signal is done iteratively by adjusting a threshold level and gradually extracting the signal present, despite the presence of artifacts from the point spread function.

Following this scheme, the signal can be recovered with high fidelity (ref, ref). The above example demonstrates the process on a 1-dimensional non-uniform signal. A similar process can be directly applied to 2 and 3-dimensional non-uniform signal.

IST reconstructions are performed by manipulating the conventional work-flow for fourier transform of regular NMR data with NMRPipe. Data is converted and transformed in the direct dimension in much the same way as regular NMRPipe. Before transform in non-uniform dimensions, the IST procedure must be applied. To do this, several scripts along with the IST software have been devised. We demonstrate examples of how to do this on this page (to the left).