1D NMR Processing in Linux and Windows Example Script

From Powers Wiki
Revision as of 03:25, 11 September 2021 by Mjeppesen (talk | contribs) (Added to Category)

Load MVAPACK

  addpath('/opt/mvapack/');

Load Data

  F.dirs = glob('???');
  [F.data, F.parms, F.t] = loadnmr(F.dirs);

Add Classes and Labels

  cls.Y = classes([7,8,6,8,8,8,7,8,7,8,8,8,8,8,8]);
  cls.labels = {'1','2','3','4','5','6','7','8','9','10','11','12','13','14','15'};

Zerofills

  F.data = zerofill(F.data, F.parms, 2); #number of zerofills can be adjusted

Build the Spectra/FT

  [S.data, S.ppm] = nmrft(F.data, F.parms);

Note: Check out the Quality (spit out a couple plots)

  #plot(S.ppm,S.data(1,:)) - Change 1 to a different number to inspect that spectrum.

Autophase the Spectra

  1. Note, some users run this ~2-3 times for convergence
  [S.data, S.phc0, S.phc1] = autophase(S.data, F.parms); 
  [S.data, S.phc0, S.phc1] = autophase(S.data, F.parms); 
  [S.data, S.phc0, S.phc1] = autophase(S.data, F.parms); 

Extract the Real Spectral Components

  XF.data = realnmr(S.data, F.parms);
  XF.ppm = S.ppm;

Check Plot to Find Reference Standard

  plot(XF.ppm, XF.data)

Reference Adjustment

  XF.ppm = refadj(XF.ppm, -0.160, 0.0);

Icoshift

  XF.data = icoshift(XF.data, XF.ppm);

Remove Undesired Regions

  r0= findnearest(XF.ppm, min(XF.ppm));
  r1=findnearest(XF.ppm, 0.4);
  r2=findnearest(XF.ppm, 0.97);
  r3=findnearest(XF.ppm, 1.33); 
  r4=findnearest(XF.ppm, 4.5);
  r5=findnearest(XF.ppm, 5.2);
  r6=findnearest(XF.ppm, 8.5);
  r7=findnearest(XF.ppm, max(XF.ppm));
  X.rm.var = [r7:r6,r5:r4,r3:r2,r1:r0];
  [XF.data, XF.ppm] = rmvar(XF.data, XF.ppm, X.rm.var);

Normalize Data

  X1.data= pqnorm(XF.data);
  X1.ppm=XF.ppm

Binning

  [B.data, B.ppm, B.widths]=binadapt(XF.data,XF.ppm,F.parms);

Build Models

Principle Component Analysis

  mdlpca= pca(X1.data); 
  mdlpca = addclasses(mdlpca, cls.Z); 
  scoresplot(mdlpca, 2, [], true);
  print -deps -color 'FB-2-mdlPCA.eps'
  print -dpdf -color 'Fb-2-mdlPCA.pdf'
  
  RQ Plot
  rqplot(mdlpca)
  print -deps -color 'FB-2-rqmdlPCA.eps'
  print -dpdf -color 'Fb-2-rqmdlPCA.pdf'
  
  Save Scores
  savescores (mdlpca,FB-2-scoresPCA,3,cls.Y,cls.labels)

Orthogonal Projection to Latent Squares (OPLS)

Note: no seperation with binned data

  mdlopls = opls(X1.data, cls.Y);
  mdlopls = addlabels(mdlopls, cls.labels);
  print -deps -color 'FB-2-mdlOPLS.eps'
  print -dpdf -color 'Fb-2-mdlOPLS.pdf'
  
  Validation Stages for OPLS Models
 
   RQ Plot
  rqplot(mdlopls);
  
  Permutation Test
  mdl.cv.perm = permtest(mdlopls);
  permscatter(mdl.cv.perm) 
 
  CV Anova
  mdl.cv.anova = cvanova(mdlopls);
  mdl.cv.anova