| Autor: |
Denise S. Weber, Warren J. Goux, George Okide |
| Rok vydání: |
1992 |
| Předmět: |
|
| Zdroj: |
Journal of Magnetic Resonance (1969). 96:526-540 |
| ISSN: |
0022-2364 |
| DOI: |
10.1016/0022-2364(92)90338-8 |
| Popis: |
Homo- and heteronuclear correlation spectroscopies have been used to completely assign the pyranosyl-ring-proton, the acetoxyl-carbonyl-carbon, and the acetoxyl-methyl-proton resonances of a variety of peracetylated α-mannosyl and β-N-acetylglucosaminyl containing di- and trisaccharide derivatives. The SIMCA class modeling algorithm and principal-component (PC) analysis were then used as a means of selecting from the data a subset of chemical shifts and coupling constants best characterizing residue types contained in new and previously assigned parent structures. It was found that between 3 and 5 of the original 15 NMR variables having the largest average interclass variances (largest discriminatory power) best described the characteristics of each residue type. Using these variable sets as a basis for the classification of residues according to the K -nearest-neighbor approach, less than 4% of the 52 “test residues” were misclassified. Both smaller and larger variable sets led to an increase in the number of misclassifications. PC plots of the data using as a basis an optimized set of variables show the potential for the classification of new datas according to the types of glycosidic linkages a residue makes to its nearest-neighbor residues. |
| Databáze: |
OpenAIRE |
| Externí odkaz: |
|
