Why is Tanimoto index an appropriate choice for fingerprint-based similarity calculations?

By Dávid Bajusz, Anita Rácz, Károly Héberger. Journal of Cheminformatics, 2015

1,525 citations recorded by the source corpus.

AbstractBackgroundCheminformaticians are equipped with a very rich toolbox when carrying out molecular similarity calculations. A large number of molecular representations exist, and there are several methods (similarity and distance metrics) to quantify the similarity of molecular representations. 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. The effects of molecular size, selection methods and data pretreatment methods on the outcome of the comparison are also assessed.ResultsA supplier database (https://mcule.com/) was used as the source of compounds for the similarity calculations in this study. 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. 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. Finally, the effects of various factors (data pretreatment, molecule size, selection method) were evaluated with analysis of variance (ANOVA).ConclusionsThis study complements previous efforts to examine and rank various metrics for molecular similarity calculations. 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. 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. these metrics could produce the rankings closest to the composite (average) ranking of the eight 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. for data fusion). Conclusions are also drawn regarding the effects of molecule size, selection method and data pretreatment on the ranking behavior of the studied metrics. Graphical AbstractA visual summary of the comparison of similarity metrics with sum of ranking differences (SRD).

Subfields: Computational Theory and Mathematics, Molecular Biology, Spectroscopy

View Paper: https://www.semanticscholar.org/paper/9afafe6095df854829d54a9172d302e272ce75a4

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computational Theory and Mathematics

1. Rohit Batra
2. Nigel W. Moriarty
3. Paul D. Adams
4. Jianzhu Ma
5. Petr Skoda
6. Jian Peng
7. Yuan Yao
8. Fei Wang

In Molecular Biology

1. Jianzhu Ma
2. Nigel W. Moriarty
3. Jinbo Xu
4. Paul D. Adams
5. Lucy J. Colwell
6. Rohit Batra
7. Jian Peng
8. Yufeng Su

In Spectroscopy

1. Jian Peng
2. Rohit Batra
3. Paul D. Adams
4. Fei Wang
5. David Hoksza
6. Jiaqi Guan
7. Petr Skoda
8. Tunca Dogan

Citations in the corpus

• Predicting protein–ligand affinity with a random matrix framework (2016, methodology): “Existing approaches then take the descriptor associated with each ligand and compare ligands with each other, for example through the Tanimoto coefficient [17, 18].”
• Emerging materials intelligence ecosystems propelled by machine learning (2020, background): “The Euclidean and Manhattan distances are not recommended, although their variability and diversity from other distance metrics might be advantageous for data fusion 75 .”
• MoFlow: An Invertible Flow Model for Generating Molecular Graphs (2020, methodology): “We choose Tanimoto index [2] as the chemical similarity metrics and indicate their similarity values by a heatmap.”
• An editor for the generation and customization of geometry restraints (2017, positive): “An additional button allows molecules that have a similar Tanimito score (Bajusz et al., 2015) to be listed using a molecular fingerprint that is a truncated version of the PubChem fingerprint.”
• 3D Equivariant Diffusion for Target-Aware Molecule Generation and Affinity Prediction (2023, positive): “…2012), synthesizability SA (Ertl and Schuffenhauer, 2009; You et al., 2018), and diversity computed as the average pairwise -5.75 -5.64 -6.18 -5.88 -6.75 -6.62 37.9% 31.0% 0 .51 0.50 0.63 0.63 0.70 0.70 -5.14 -4.70 -6.42 -5.82 -7.15 -6.79 Tanimoto distances (Bajusz et al., 2015; Tanimoto, 1958).”
• Controlling an organic synthesis robot with machine learning to search for new reactivity (2018, methodology): “Under basic conditions, phenylacetyl chloride ( 15 ) is deprotonated by DBU, giving phenyl ketene, which reacts with DBU to give the polycyclic azepine derivative 22 (Fig.”
• A 3D Generative Model for Structure-Based Drug Design (2022, methodology): “Metrics We assess the generated molecules from fragments with four main metrics: (1) Similarity : We use Tanimoto Similarity [32, 3] over Morgan fingerprints [14] to measure the similarity between the molecular graphs of generated molecule and the reference molecule.”
• Benchmarking platform for ligand-based virtual screening (2016, methodology): “up analysis step is introduced, for example, principal component analysis [12] or statistical methods [13] can be employed to gain additional insight.”
• Pocket2Mol: Efficient Molecular Sampling Based on 3D Protein Pockets (2022, positive): “Train stands for the Tanimoto similarity (Bajusz et al., 2015) with the most similar molecules in the training set; (8) Diversity is calculated as the average Tanimoto similarities of the generated molecules for the individual pockets; (9) Time represents the time cost to generate 100 valid…”
• Therapeutic target database 2020: enriched resource for facilitating research and early development of targeted therapeutics (2019, positive): “Moreover, 3388 of the 5,059 agents include experimental target binding activity data and 2215 of the 5059 agents were mapped with PubChem IDs (11) according to a structural search (Tanimoto coefficient (64) between molecular fingerprints equals to one) and a subsequent visual inspection of all match”
• Recent applications of deep learning and machine intelligence on in silico drug discovery: methods, tools and databases (2018, background): “Various types of measures have been proposed for this purpose such as the Dice coefficient [74] or the Tanimoto coefficient, which currently is the most popular similarity measure for compounds [62, 75].”
• Milvus: A Purpose-Built Vector Data Management System (2021, background): “, with Tanimoto distance [9]) to find similar structures [9, 66].”
• Large-Scale Off-Target Identification Using Fast and Accurate Dual Regularized One-Class Collaborative Filtering and Its Application to Drug Repurposing (2016, background): “Our result supports a recent study [25] which showed that Tanimoto coefficientis efficient for the chemical similarity calculation.”