Compound molecular property predictor

It makes sense to use data mining of structural features and molecular properties to assist in the building of alerts for more precise structural features. Various in vitro testing results can be used to describe structures at the compound and feature levels. These profiles can be used either to build weight of evidence strategies or as predictors for prediction models. 

Example In molecular structure elucidation the observations are pairs of spectra and compounds. Predictors used are spectral predictors, functions that map spectra onto real numbers. The target variable is, for example, a binary molecular descriptor of a structural property SP, equal to 1 if a compound has property SP, and equal to 0 otherwise. The search is for a function able to predict whether or not the corresponding compound has property SP for a given spectrum. We will calculate such predicting functions in Section 8.5. 

The genesis of in silico oral bioavailability predictions can be traced back to Lip-inski s Rule of Five and others qualitative attempts to describe drug-like molecules [13-15]. These processes are useful primarily as a qualitative tool in the early stage library design and in the candidate selection. Despite its large number of falsepositive results, Lipinski s Rule of Five has come into wide use as a qualitative tool to help the chemist design bioavailable compounds. It was concluded that compounds are most likely to have poor absorption when the molecular weight is >500, the calculated octan-l-ol/water partition coefficient (c log P) is >5, the number of H-bond donors is >5, and the number of H-bond acceptors is >10. Computation of these properties is now available as an ADME (absorption, distribution, metabolism, excretion) screen in commercial software such as Tsar (from Accelrys). The rule-of-5 should be seen as a qualitative, rather than quantitative, predictor of absorption and permeability [16, 17]. 

The study of the function y = f x) starts from measurements of the properties of many compounds, and subsequent compilations into tables and databases. An analysis of these empirical observations can lead to useful associations and trends, and generalizations that may have predictive power. When a sufficiently large and systematic database has been accumulated, researchers will try to find correlations between a property y and predictors, which are parameters relating to molecular stmcture x, or other more easily available properties, y". 

In contrast to a chemical property which can be measured, a molecular descriptor is computed from the molecular structure. Contained in the structural information are the atoms making up the molecule and their spatial arrangement. From the coordinates of the atoms, the geometric attributes (i.e., the size and shape of the molecule) can be deduced. A straightforward example is the molecular mass, which is computed by adding up the masses of the individual atoms making up the molecule and indicated in the elemental composition. The result is accurate since the atomic masses are independent of the chemical bonds with which they are involved. However, the molecular mass reflects few of the geometrical and chemical attributes of a compound and M is therefore a poor predictor for most properties. 

Common chemometric tools may be applied to deal with similarity matrices. Particularly, partial least squares (PLS) [73,74] stands as an ideal technique for obtaining a generalized regression for modeling the association between the matrices X (descriptors) and Y (responses). In computational chemistry, its main use is to model the relationship between computed variables, which together characterize the structural variation of a set of N compounds and any property of interest measured on those N substances [75-77]. This variation of the molecular skeleton is condensed into the matrix X, whereas the analyzed properties are recorded into Y. In PLS, the matrix X is commonly built up from nonindependent data, as it usually has more columns than rows hence it is not called the independent matrix, but predictor or descriptor matrix. A good review, as well as its practical application in QSAR, is found in Ref. 78 and a detailed tutorial in Ref. 79. 

---