Probability-based matching system

Pesyna, G. M., Venkataraghavan, R., Dajuinger, H. E., McLafferty, F.W. Probability-based Matching System Using a Large Collection of Reference Mass Spectra. Anal. Chem. 1976,52, 1362-1368. 

Atwater, B.L., Stauffer, D.B., McLafferty, F.W., and Peterson, D.W. (1985) Reliability Ranking and Scaling Improvements to the Probability Based Matching System for Unknown Mass Spectra. /ItiaL Chem., 57, 899—903. 

Three different mass spectrometry search algorithms dominate the database searching systems commercially available today. The Cornell University Probability Based Matching (PBM) software, The Integrated Control System (INCOS) and the MassLib system (see below) with the SISCOM search software. 

A completely different search strategy forms the basis of the PBM algorithm (probability based match). The statistical mathematical treatment by Prof. McLafferty allows predictions to be made on the probable identity of a substance suggestion (Atwater et al, 1985 Palisade Corporation, 1994). The search procedure was developed in the 1970s at Cornell University as part of the Cornell algorithm (STIRS, the self-training interpretative and retrieval system as an interpretative system). In the subsequent years, parts of the PBM procedure... 

Forensic laboratories in the United States have agreed on 13 core STR loci that are most accurate for identification of an individual. Based on these 13 loci, a Combined DNA Index System (CODIS) has been established to serve as a registry of convicted offenders. If the profile of sequences from a known individual and the profile from DNA obtained at a crime scene match, the probability is approximately 82 billion to 1 that the DNA is from the same individual. In paternity cases, where the DNA of father and offspring are related but not fully identical, the identity of the father can be established with a probability of 100,000 to 1. 

The investigation of failures of manufactured components and systems, especially in the electronics and aerospace industries, has generated a variety of statistical models on which data analysis may be based. Each model uses a specific distibution of failure probabilities, and it is important to select a model that matches the actual distribution inherent in the product concerned. In the case of dielectric breakdown, where a large number of quite different modes of failure are known to occur, sometimes even together, the application of a particular statistical failure model must be approached with great caution. Nevertheless, one treatment, based on a Weibull distribution of failure probability, has taken root, and is most generally used in practice. For a dielectric, the Weibull failure probability function has the form... 

Statistical mechanical approaches apply mainly to deductions about structure and are the basis of interpretations of the entropy of ions in solution and the solution s heat capacity. The entropy of a system can be calculated if the partition functions of the ions and the water molecules surrounding them are known." The partition functions (translatory, rotational, and vibrational) can be obtained from textbook material by assuming a structure of the ion-solvent complex. By comparing calculations based on various assumptions about stmcture with the values obtained from experiments, certain stmctures can be shown to be more likely (those giving rise to the calculations that match the experiment), others less probable, and some so far from the experimental values that they may be regarded as impossible. 

The need to identify microorganisms rapidly has led to the development of a number of molecular identification and characterization tools. These have not yet become routinely adopted in the analytical or diagnostic laboratory but will probably do so in the future. One such technique isolates and amplifies 16S ribosomal DNA and, following sequencing of the bases, compares this with known sequences held in a reference library. This approach enables phylogenetic relationships to be derived even for those bacteria that have not previously been identified. Other systems examine the patterns of key constituents of the cells such as fatty acids and assign identities based on similarity matches to known reference cultures. 

Statistical models for the analysis of NMR data are used in two complementary approaches (Fig. 2) an analytical (model fitting) approach and a synthetic (computer simulation) approach. In the analytical approach, assigned NMR resonance intensities are fit to expected intensities based on statistical models. In the synthetic approach, spectral intensities are first calculated using reaction probabilities predicted by theoretical models these theoretical intensities are matched with those observed in the NMR spectrum. The calculation is based on theoretical probability expressions or Monte Carlo simulation. In an integrated approach, both methods are used for more complex systems. 

---