Similarity-based method

Lajiness MS. Molecular similarity-based methods for selecting compounds for screening. In Rouvray DH, editor, Computational chemical graph theory. New York Nova Science Publishers, 1990 299-316. 

Similarity is a fundamental concept that has been used since before the time of Aristotle. Even in the sciences, it has been used for more than two centuries (1). Similarity is subjective and relies upon comparative judgments—there is no absolute standard of similarity, rather like beauty, it is in the eye of the beholder. Because of this subjectivity it is difficult to develop methods for unambiguously computing the similarities of large sets of molecules (2). Moreover, there is no absolute standard to compare to so that assessing the validity of any similarity-based method remains subjective basically, one must rely upon the... 

An ROC curve plots the true positives against the false positives for different classifications of the same set of objects this corresponds to plotting a against n - a using the notation of Table 1, and thus the shape of an ROC curve tends to the shape of a cumulative recall plot when n a. An example of the use of ROC plots in chemoinformatics is provided by the work of Cuissart et al. on similarity-based methods for the prediction of biodegradability (24). 

In this chapter we have discussed the foundations of molecular similarity analysis that are largely responsible for success or failure of similarity-based methods in the context of virtual compound screening. Furthermore, we have described how crucial underlying SAR characteristics are for molecular similarity analysis and presented a comprehensive methodological framework for the qualitative and quantitative analysis of SARs and the classification of different SAR types. 

Lajiness, M.S. (1990). Molecular Similarity-Based Methods for Selecting Compounds for Screening. In Computational Chemical Graph Theory (Rouvray, D.E., ed.). Nova Science Publishers, New York (NY), pp. 299-316. 

Knowledge or similarity-based methods exploit similarities between a given sequence and (sequences of) proteins of known 3D structure the underlying assumption is that the more similar the two sequences, the more similar are the two corresponding 3D structures. The similarity is often generalized to additionally include quasi-physical properties, such as preferences of amino acids to occur in certain states, e.g. buried inside the protein or... 

Similarity is very subjective, as it depends on what are we looking for and from what point of view we are looking. For example, from a mathematical point of view, we would denote two molecules as similar if they have common features in three dimensions, whereas if we take a chemical approach, we would denote two molecules as similar if they had similar physical properties. Similarity-based methods have gained popularity due to the rapid technological progress and increased number of entries in chemical databases. This has made the application of computational search methods a necessity. 

A serious limitation of all similarity-based methods is that chemical spaces are not invariant to the representation (or similarity function) used. Thus, nearest-neighbors in one space may not be nearest-neighbors in another [41], and what is similar to an active compound in one space (hence also potentially active) may not be in another, begging the question of which space should be used. In addition, the presence of activity cliffs [152-156], which occur when small changes in structure are accompanied by large changes in activity, can further confound the application of similarity methods since they violate the similarity-property principle [22] leading to what Stahura and Bajorath [157] have termed the similarity paradox. ... 

Baraldi, P, Di Maio, F., Genini, D. and Zio E., 2013, A fuzzy similarity based method for signal reconstruction during plant transients, Chemical Engineering Transactions, 33, pp. 889 894. 

---