Processing 3D HNCACB data.

• The data is processed through nmrPipe.
• Planes from the data (which will be carbon by hydrogen) will be examined by nmrDraw to see if spots from the HSQC are represented by columns with 2 positive and 2 negative signals in the carbon dimension.  I guess particular attention should be paid to seeing if the HSQC signals with known marginal T2 times are making it through. If not, we need to fall back to HNCA and an HN(CA)CB determinations.  If it looks OK, the CBCACONH should be started.  If CBCACONH is too insensitive, it is possible to break it to a HN(CO)CA and HN(COCA)CB.
• Then a program: plotpseq is used to convert the 3D data to strips (essentially a 2D array of columns centered on HSQC signals).  The input appears to be derived from a table of picked peak coordinates from the HSQC.  There is a web page with some instructions linked off of Andy's unix software page.
• The strips can be viewed with nmrDraw or PIPP.
• Each 4 spot series gives C alpha and beta values for this residue and the adjacent one.  Therefore strips likely to be adjacent to each other in the sequence are identified.  There is some reciprocol intesity relationship that is also used.
• Likely identity of residues is inferred from a table of the distribution of C alpha and C beta resonance frequencies of  the various residues.  The table of standard values is obtained from biomagresbank.  It's retrieved by clicking the <retrieve> button on the upper left of the page and looking for the relevant tabulation in that menu.  Glycine of course will have no C beta signals.
• Then trick is to match the ambiguous subsequences to the real sequence.  Linkage won't be possible through proline, right?
• Andy says that there are a variety of different strategies for organizing the task, including using some programs that may or may not be trustworthy.  I'm thinking to let Perl's regular expression matching function do it.
• I'll have a substantial segment that's disordered.  They'll be obvious because they are so bright.  Andy thinks to use the same table of characteristic resonances as for the structured regions.  I also notice a table at BioMagResBank with value from pentapeptides.
• I should also have a minority signal from a unfolded portion of the folded part of the protein.  These signals may be identifiable due to being narrower.  Actually they probably won't be a minority signal after the T2 effect beats down the signal from the folded portion of the protein.  In principle, I ought to be able to assign a random coil all the way through the sequence, and then tell which of the residues are also in the folded portion by the much diminished intensity of the random coil signal in the HSQC at those positions.  The center part of the proton spectrum where all this happens might be pretty crowded though.



---