Recognition of Molecular Structures

The most chal lenging task of pattern recognition applications in mass spectrometry is the automat i c recogni t i on of molecular subst ruetures in a molecule. In a g reat number of papers c Lassi fi ers for this purpose were developed by the learning machine method typical predictive abilities are 70 to 95 %. 

Modi fi cat ions of the learni ng machine, appropri ate preprocessing, and feature selection improved the classification results. Use of cross terms (whi ch take into account interact i ons between two mass numbers) accelerated the training but had less influence on the predi ctive abilities C1243- The introduction of a width parameter into the learning machine slightly improved the predictive ability and the absolute value of the scalar product could be used as a measure of confidence C3203. 

A Fourier transformation of the mass spectra had almost no effect on the predictive abilities C122, 3213. 

Combination of the mass spectrum, infrared spectrum, melting point, and boiling point to a single pattern vector improved the results for the recognition of a carbon-carbon double bond, but had almost no influence on the recognition of an ethyl or vinyl group C1303. 

Interpretation of mass spectra by distance measurement to centres of gravity has been reported by Justice and Isenhour C133, 1353, Mathews 

---

Apart from detail, reformulation of quantum theory to be consistent with chemical behaviour, requires the recognition of molecular structure. In this spirit, it may be introduced as an essential assumption, or emergent property, without immediate expectation of retrieving the concept from first principles. Medium-sized molecules, especially in condensed phases, are assumed to have a characteristic three-dimensional distribution of atoms, which defines a semi-rigid, flexible molecular frame. The forces between the atoms are of quantum-mechanical origin, but on a macro scale, are best described in terms of classical forces. 

The search for optimal and near-optimal solutions is an important problem in different areas of computational chemistry. This could involve the search for molecules and/or structures possessing desired activities and properties, the development of optimal models for prediction of activity of classes of compounds, recognition of molecular structures from spectroscopic data or in databases, prediction of active conformers for proteins and minimization of energy potentials in combinatorial chemistry. The traditional paradigm for... 

The molecular scaffolds also provide an unusual opportunity for the identification and recognition of meso structures. For example, successive acylation of L and D phenylalanine with the acridine diacid gives the structure 56 and we anticipate that... 

Two pattern recognition techniques are applied to the analysis of the library of FTIR spectra compiled by the US EFA> The patterns which emerge demonstrate the influence of molecular structure on the spectra in a way familiar to chemical spectroscopists They are also useful in evaluation of the library, which is not error free, and in assessing the difficulties to be expected when using FTIR spectra for complex mixture analysis. 

T.R. Stouch and P.C. Jurs, Computer-Assisted Studies of Molecular Structure and Genotoxic Activity by Pattern Recognition Techniques,... 

The structural analysis has been carried out right up to the recognition of molecular Level (i, 2). a-Chymotrypsin is poly(amino acid) consisting of 245 amino adds, having relatively deep grooves. It catalyzes the hydrolysis of carboxylic acid derivatives such as protein, simple amides, esters, etc. The active site is composed of aspartic add, Asp (102). .. histidine, His (57). .. serine, Ser (195), and the distances between Asp. .. His and His... Ser are 2.8 A and 3.0 A, respectively. Electronic structures of these moieties depend on the pH of the reaction system. In the range of pH > 7 at which a-chymotrypsin is active, —COO" of Asp attracts N4 proton in imidazolyl of His, and Nj in the imidazolyl of His attracts the proton in OH of Ser. It is called charge-relay system . 

Small structural changes are especially important in molecular recognition. It has been noted, e.g., that ... subtle changes of molecular structure may result in severe changes of inclusion behavior of a potential host molecule due to the complicated interplay of weak inter-molecular forces that govern host-guest complex formation [41],... 

Geometrical considerations have gained additional importance due to their role in molecular recognition which implies the (molecular) storage and (supramolecular) retrieval of molecular structural information [82], The formation of supramolecular structures necessitates commensurable and compatible geometries of the... 

ADAPT Pattern recognition Cluster analysis Uses QSAR descriptors of molecular structure Limited to congeneric series of chemicals... 

For many years the focus of a large number of theoretical chemists was on the development of better computational methods and approximations. There was a small, but growing number of chemists who were interested in applying these methods to practical questions of concern to experimentalists. Computational chemistry has now achieved de facto recognition as a discipline through formation of the Division of Computers in Chemistry of the American Chemical Society and through the publications of Journal of Computational Chemistry, Journal of Molecular Structure, Computers and Chemistry, and Journal of Molecular Graphics. 

The ordered sequence of the n smallest eigenvalues of the B matrix was proposed as a molecular descriptor with high discrimination power, to be used in the recognition and ordering of molecular structures. The basic assumption was that the lowest eigenvalues contain contributions from all atoms and thus reflect the topology of the whole molecule. 

It might thus be expected that this structure has been incorporated in many other proteins wherever there was a need for its characteristic function. A requirement of binding nucleotides to proteins occurs in energy transfer systems and in the molecular reproduction of nucleic acids. Consequently, it is possible that this molecular fossil may be found in such diverse proteins as tRNA synthetase, ribosomal proteins, and virus coat proteins. The recognition of its structure by sequence homology or from X-ray structure determinations may also give guidance as to function where none is known. 

---