! /buddha/rwb1002/samples/README_baseline

R.W.Broadhurst, Department of Biochemistry, University of Cambridge, U.K.
13/Oct/1994

The best way to get a flat baseline in a spectrum is to set up
the experiment carefully in the first place.  Good initial data
is always better than bad data that has been massaged by
extensive processing.  Experiments can be set up so that they
need a (0,0) or (90,0) phase correction in the acquisition
dimension d1 by modifying the receiver phase, and it is sometimes
necessary to adjust the pre-acqusition delay so that only a linear
baseline correction is required. The phase correction
for indirect dimensions can be set, for example, to (90, -180) or
(180, -360) by correct choice of the initial value of the 
incremented delay.  A flat baseline in the indirect dimension
also tends to be a by product.   This method is outlined in

BAX_A, IKURA_M, KAY_LE & ZHU_G, J MAG RESON 91, 174 (1991)
	(for States)
ARCHER_SJ, BALDISSERI_DM & TORCHIA_DA, J MAG RESON 97, 602 (1992)
	(for TPPI)

Further suggestions about experimental methods are in the literature,

eg. WALTHO_JP & CAVANAGH_J, J MAG RESON, 103A, 338 (1993).

However, even if these procedures have been followed, some problems
may remain, especially in samples run with presaturation on samples
in H2O. 

In some cases (eg. experiments involving proton-hereronuclear
correlations like an HSQC) the spectra are greatly improved by
deconvolution of the low frequency water signal.  The "process"
program can do this using the "convolve" command.  The same
technique might also improve a 2D NOESY or TOCSY spectrum, but
will have two disadvantages.  First, signal near zero frequency
in d1 will be lost.  Second, if the water signal is intense,
the d1 baseline may have the form of a postive (or negative) 
ridge with a U-shaped dip (or n-shaped hill) to zero intenisty
centered at zero frequency.  Further baseline correction will
be necessary.

If the spectrum is well behaved, a single baseline correction
function will be adequate for the complete spectral width.
In this case it is best to use the commands "base_const",
"base_poly" or "base_trig".  In there cases it is convenient
to do the baseline correction at the end of the processing
command list for each dimension. 
eg. performing the d2 baseline correction with the command
	
	base_const 8

in the command list /buddha/rwb1002/samples/amx/hmqc/rwb18_7_d2.com

If there are problems associated with an intense signal,
eg. the water signal in d1, giving a difficult baseline
function to fit, then it is wise to baseline correct
different portions of the baseline separately, using the 
commands "base_const2", "base_poly2" or "base_trig2".  So,
in d1 we could correct the baseline on either side of the
water signal.
eg. using the commands

base_trig2 2 2 1 950            ! Baseline correction: different functions
base_poly2 2 2 1100 2048        ! can be applied either side of the H2O
                                ! signal

in /buddha/rwb1002/samples/am/noesy/tm9e5a_d1.com

If the spectrum is really difficult, some improvement can be made
by using smaller baseline strips

eg. 

base_poly2 4 1 1 430
base_trig2 4 2 431 650
base_poly2 4 1 651 800
base_poly2 4 1 801 900
base_poly2 4 1 901 948
base_poly2 4 1 949 976
base_poly2 4 2 977 1000
base_poly2 4 2 1001 1036
base_poly2 4 2 1037 1100
base_poly2 4 1 1101 1300
base_poly2 4 1 1301 1500
base_poly2 4 1 1501 2048

In these cases better results often come by applying the d1 baseline
correction after the whole data set has been processed and calling
a baseline correction command list in the processing control script 
file (ie. a file like tm9e5a_2dft.scr). eg.

script 1 tm9e6a_d1.com          ! Command list for d1 processing (FFT)
script 2 tm9e6a_d2.com          ! Command list for d2 processing (FFT)
script 1 tm9e6a_d1a.com		! Command list for d1 baseline correction

If much adjustment is needed, it might be worthwhile saving the spectrum 
after each processing stage so that the baseline correction command list
can be changed and the results compared in the "plot2" program.

