Sequence analysis of oligodeoxyribonucleotides by mass spectrometry. 2. Application of computerized pattern recognition to sequence determination of di-, tri-, and tetranucleotides.

A novel strategy for the sequence analysis of oligodeoxyribonucleotides has been devised which is based upon the analysis of intact underivatized oligonucleotides by mass spectrometry followed by interpretation of the mass-spectral data by computerized pattern-recognition techniques. The pyrolytic and electron-impact conditions of the mass spectrometer permit the cleavage of oligonucleotides of varying chain length and composition, yielding reproducible fragmentations and characteristic m/e values which can be used to reveal purine and/or pyrimidine base sequence information. The selection of optimum features (which are the ratios of peak heights of specific ions, or the linear combination of such ratios) has been done by an interactive feature selection method employing multidimensional k nearest-neighbor analysis and two-dimensional feature-space plots (nonlinear mappings) of the mass-spectral data. Features have been found which allow 100% classification accuracy in predicting the 5' and 3' terminus of all of the dinucleotides commonly found in DNA. Other specific features have been found which indicate adjacent nucleotides within a tetranucleotide. Knowledge of the adjacent nucleotide pairs present, in conjunction with the information as to the 3' or 5' position of the residues in each pair, permits the reconstruction of the sequence of the tetranucleotide.