Decoys Selection in Benchmarking Datasets: Overview and Perspectives

Abstract

Virtual Screening (VS) is designed to prospectively help identifying potential hits, i.e., compounds capable of interacting with a given target and potentially modulate its activity, out of large compound collections. Among the variety of methodologies, it is crucial to select the protocol that is the most adapted to the query/target system under study and that yields the most reliable output. To this aim, the performance of VS methods is commonly evaluated and compared by computing their ability to retrieve active compounds in benchmarking datasets. The benchmarking datasets contain a subset of known active compounds together with a subset of decoys, i.e., assumed non-active molecules. The composition of both the active and the decoy compounds subsets is critical to limit the biases in the evaluation of the VS methods. In this review, we focus on the selection of decoy compounds that has considerably changed over the years, from randomly selected compounds to highly customized or experimentally validated negative compounds. We first outline the evolution of decoys selection in benchmarking databases as well as current benchmarking databases that tend to minimize the introduction of biases, and secondly, we propose recommendations for the selection and the design of benchmarking datasets.

Computational chemistry in drug lead discovery and design

“…Ref. [] for review). The use of polar surface area as an alternative charge‐related physical property was also evaluated, but the use of the formal charge descriptor was preferred, mainly due to practical reasons.…”

“…Chemical diversity should be sought to maximize the probability of finding novel ligands. In case of retrospective molecular docking, availability of ligand binding information and an un‐biased database of decoys (small‐molecules which are assumed to be non‐binders) would help to obtain reliable results from HTD The docking strategy: In HTD, two different stages can be identified for each molecule: (1) the docking stage, where an accurate pose (or low‐energy poses) of the molecule within the binding site is sought through optimizing protein‐molecule interactions; (2) the scoring stage, where low energy poses (or the lowest one) are assigned a score aimed as a measure of the probability of that molecule to actually bind to the target.…”

Ligand‐ and Structure‐Based Approaches of Escherichia coli FabI Inhibition by Triclosan Derivatives: From Chemical Similarity to Protein Dynamics Influence

“…pointed out that the assignment of different weights to the chemical features (represented by different colours), therefore changing the combo score, could improve the performance in virtual screening . Additionally, models’ enrichment can either be over or underestimated by inappropriate decoy selection . Specifically, DUD‐E's decoy set was shown to overestimate AUC values when used to validate machine learning‐based models .…”

“…As CSMs calculate the similarity between a compound and a query defined by the molecular volume and some chemical features distributed in specific positions on space, we assumed that dissimilar compounds in decoy set may not interfere in the model's evaluation. Accordingly, the lack of chemical diversity in active compounds set is a major limiting step for the model training and can become a source of bias to the model . Then, for this work, all analysed subsets (active, inactive and decoys compounds) were similar in terms of physicochemical properties to avoid bias.…”

MUBD‐DecoyMaker 2.0: A Python GUI Application to Generate Maximal Unbiased Benchmarking Data Sets for Virtual Drug Screening

“…Since the release of the first benchmarking sets, constant efforts have been taken to improve the quality of benchmarking sets by reducing three main types of biases, i. e. artificial enrichment, analogue bias and false negative bias . Among the currently available benchmarking sets, MUV, DEKOIS, DUD‐E, NRLiSt BDB and MUBD‐HDACs were selected as the state‐of‐the‐art as of 2017 . It is worth to mention that MUBD‐HDACs, with the full name of Maximal Unbiased Benchmarking Data sets for HDACs, was made possible from our research effort.…”