Hit Dexter 2.0: Machine-Learning Models for the Prediction of Frequent Hitters

Abstract

Assay interference caused by small molecules continues to pose a significant challenge for early drug discovery. A number of rule-based and similarity-based approaches have been derived that allow the flagging of potentially "badly behaving compounds", "bad actors" or "nuisance compounds". These compounds are typically aggregators, reactive compounds and/or pan-assay interference compounds (PAINS), and many of them are frequent hitters. Hit Dexter is a recently introduced machine learning approach that predicts frequent hitters independent of the underlying physicochemical mechanisms (including also the binding of compounds based on "privileged scaffolds" to multiple binding sites). Here we report on the development of a second generation of machine learning models which now cover both primary screening assays and confirmatory dose-response assays. Protein sequence clustering was newly introduced to minimize the overrepresentation of structurally and functionally related proteins. The models correctly classified compounds of large independent test sets as (highly) promiscuous or nonpromiscuous with Matthews correlation coefficient (MCC) values of up to 0.64 and area under the receiver operating characteristic curve (AUC) values of up to 0.96. The models were also utilized to characterize sets of compounds with specific biological and physicochemical properties, such as dark chemical matter, aggregators, compounds from a high-throughput screening library, drug-like compounds, approved drugs, potential PAINS and natural products.Among the most interesting outcomes is that the new Hit Dexter models predict the presence of large fractions of (highly) promiscuous compounds among approved drugs. Importantly, predictions of the individual Hit Dexter models are generally in good agreement and consistent with those of Badapple, an established statistical model for the prediction of frequent hitters. The new Hit Dexter 2.0 web service, available at http://hitdexter2.zbh.uni-hamburg.de, not only provides user-friendly access to all machine learning models presented in this work but also to similarity-based methods for the prediction of aggregators and dark chemical matter as well as a comprehensive collection of available rule sets for flagging frequent hitters and compounds including undesired substructures.

Veratrilla baillonii Franch Could Alleviate Lipid Accumulation in LO2 Cells by Regulating Oxidative, Inflammatory, and Lipid Metabolic Signaling Pathways

“…The PubChem ( ) and HIT ( ) databases were utilized to identify the verified targets of each active component ( Song et al., 2018 ). In order to identify the potential targets of V. baillonii , the molecular similarity match tool was used based on the simplified molecular input line entry specification (SMILES) in the similarity ensemble approach (SEA) ( P <0.05) ( ) and SwissTargetPrediction ( P <0.05) ( ) ( Stork et al., 2019 ). The UniProt ( ) database was used to standardize the results.…”

“…bkslab.org/) and SwissTargetPrediction (P<0.05) (http://www. swisstargetprediction.ch/) (Stork et al, 2019). The UniProt (https://www.uniprot.org/) database was used to standardize the results.…”

Applications of Deep-Learning in Exploiting Large-Scale and Heterogeneous Compound Data in Industrial Pharmaceutical Research

“…Thus, these filters should be applied with care (Baell and Nissink, 2018). Stork et al (2018, 2019) developed the Hit Dexter model to predict frequent-hitter, aggregator, PAINS, dark chemical matter (Wassermann et al, 2015b), and other potential nuisance compounds. The Hit Dexter model is based on a set of extensively tested compounds from PubChem represented by their 2D molecular fingerprints.…”

Identification of Compounds That Interfere with High‐Throughput Screening Assay Technologies

“…Most published computational promiscuity analysis methods rely on statistical analysiso fh istorical activity data extracted from primary screens [17,20,25] and, lessfrequently,f rom confirmatory concentration-response screens. [18,26] An unusually high hit rate for ac ompound across many screens for an HTS technology can be an indication that the compound is causingi nterference. Similarly,i facompound is active against aw ide range of targets or target families, it mustb et reated carefully as it is likely unselective or perhaps prone to false readouts.…”

“…While chemical scaffolds can be inherently promiscuous, [13,14] any specific pharmacophore or substructure can have ah igh impact on ac ompound's promiscuity. [15,16] The Hit Dexter platform [17,18] predicts various types of frequent-hitters, which are compounds that are active in many assays,b ased on as et of highly tested PubChem compounds represented as molecular fingerprints. Hit Dexter 2.0 covers both primary andc onfirmatory dose-response assays and classifies compoundsa sp romiscuous or not with an MCC of 0.64 and aR OC AUC of 0.96.…”