Introduction

While I hired on to work on a completely different set of projects during graduate school at UNL, I quickly gravitated towards writing software solutions for the other members of the group, whose main research goals were founded on the techniques of NMR metabolomics.

The MVAPACK toolbox originally began as a small set of utilities I could use to rapidly load 1D ¹H NMR spectra into a GNU Octave environment for whatever mathematical manipulation I was working on at the time. Once I saw how simple it could be to manipulate NMR datasets as matrices in Octave, I began writing more support code for NMR processing. Finally, the MVAPACK suite had arrived when I implemented multivariate analysis routines, making the code a one-stop shop for 1D NMR chemometrics.

Before you tell me that it’s been done before: I know. But this is the first toolbox of it’s kind that’s free and open-source.

An example pipeline

We work on Bruker spectrometers, so I implemented routines that can take raw Bruker Free Induction Decay data files into octave and process them under full automation, taking the data all the way to the final multivariate models with little to no human intervention.

Here are a few figures I captured as an example of the path the data takes through the MVAPACK software for a general metabolomics experiment. I just bought coffee downtown, lyophilized it, resuspended in heavy water, and ran the NMR. :)

The code

The official MVAPACK sources are freely available for academic use at my graduate research group’s website.

• Official MVAPACK page
• Publication on MVAPACK

MVAPACK was recently upgraded to support multiblock bilinear decompositions, including MB-PCA, MB-PLS and the newly developed MB-OPLS. You can read about the use of multiblock methods in MVAPACK here:

• Publication on MB-PCA/MB-PLS
• Publication on MB-OPLS