![]() ECFP4 and ECFP6 performed equally well, the notations are explained in the corresponding reference). Based on their evaluation with the calculation of retrieval rates of active molecules, extended connectivity fingerprints performed best (although only slightly better from the runner-up SEFP4, LCFP4 and FCFP4/6 fingerprints), regardless of diameter ( i.e. However, using different fingerprints, different (active) molecules were retrieved, which suggests the use of orthogonal fingerprints individually in virtual screenings. fingerprints that capture different aspects of molecular structure) could be identified, the use of multiple fingerprints for consensus scoring only marginally improved the results obtained with a single fingerprint. Despite the fact that diverse fingerprints ( i.e. They were able to estimate the extent to which the information captured by these descriptors overlap, and also to visualize them in a three-dimensional space. ![]() Previous work aiming to compare and assess such methods includes a 2009 article by Bender and coworkers, in which 37 molecular fingerprints were compared and their similarities were quantified (based on their rank-orderings of the same dataset) by means of statistical methods, such as principal component analysis (PCA). Even though much effort has been made to reveal and assess numerous possibilities, our knowledge is still relatively scarce about the effects the choice of methods has on the outcome of molecular similarity calculations and rankings. Meanwhile, a virtually infinite “method space” is available and waiting to be explored, with a plethora of molecular representations and a significant number of similarity (or conversely, distance) definitions to compare these representations. Although some commonly applied best practices for molecular similarity calculations exist, they are mostly based on practical experience. Its applications encompass a number of fields, mostly medicinal chemistry-related, such as virtual screening. Quantifying the similarity of two molecules is a key concept and a routine task in cheminformatics. Conclusions are also drawn regarding the effects of molecule size, selection method and data pretreatment on the ranking behavior of the studied metrics. The similarity metrics derived from Euclidean and Manhattan distances are not recommended on their own, although their variability and diversity from other similarity metrics might be advantageous in certain cases ( e.g. these metrics could produce the rankings closest to the composite (average) ranking of the eight metrics. The Tanimoto index, Dice index, Cosine coefficient and Soergel distance were identified to be the best (and in some sense equivalent) metrics for similarity calculations, i.e. ![]() Here, however, an entirely general approach was taken to neglect any a priori knowledge on the compounds involved, as well as any bias introduced by examining only one or a few specific scenarios. This study complements previous efforts to examine and rank various metrics for molecular similarity calculations. Finally, the effects of various factors (data pretreatment, molecule size, selection method) were evaluated with analysis of variance (ANOVA). Similarity metrics were compared based on their ranking of the compounds within one experiment (one dataset) using sum of ranking differences (SRD), while the results of the entire set of experiments were summarized on box and whisker plots. A large number of datasets, each consisting of one hundred compounds, were compiled, molecular fingerprints were generated and similarity values between a randomly chosen reference compound and the rest were calculated for each dataset. ResultsĪ supplier database ( ) was used as the source of compounds for the similarity calculations in this study. The effects of molecular size, selection methods and data pretreatment methods on the outcome of the comparison are also assessed. In this work, eight well-known similarity/distance metrics are compared on a large dataset of molecular fingerprints with sum of ranking differences (SRD) and ANOVA analysis. A large number of molecular representations exist, and there are several methods (similarity and distance metrics) to quantify the similarity of molecular representations. Cheminformaticians are equipped with a very rich toolbox when carrying out molecular similarity calculations.
0 Comments
Leave a Reply. |