Thus, they are powerful to identify molecules derived from the same chemical series. Historically, fingerprint similarity measures have been developed to classify molecules into families. Standard approaches use similarity measures between molecules based on chemical fingerprints ( Dunkel et al., 2008 Keiser et al., 2009 Wang et al., 2013). A widely used strategy consists in identifying proteins with known ligands similar to a query molecule (i.e. More generally, one can expect that even a significant fraction of Food and Drug Administration (FDA)-approved drugs have at least some unknown target.Ĭomputational predictions of bioactive molecule targets are helpful to narrow down the set of potential targets to be tested and to predict off-target effects of known molecules or drugs ( Keiser et al., 2009 Kuhn et al., 2013 Lounkine et al., 2012). Later on, it was shown to also bind hydroxytryptamine receptors ( Keiser et al., 2009). For instance, N, N-dimethyltryptamine was initially described as a ligand of sigma-1 receptor ( Fontanilla et al., 2009). Moreover, even for well-studied molecules, our knowledge of their targets is far from complete. For instance, 17.4% of the compounds in ChEMBL with reported functional activity in human cells do not have direct target information (see Methods Section 2.1). This is especially true for compounds tested uniquely in functional assays. However, a significant fraction of bioactive molecules still do not have any known target. annotated as binding) with activity 200 000 small molecules. For instance, only for human protein ligands, the ChEMBL database ( Gaulton et al., 2012) contains close to 350 000 reported direct interactions (i.e. These databases contain unprecedentedly large datasets of interactions between proteins and small molecules. Several databases have been developed by various groups to provide access to these data, such as ChEMBL ( Gaulton et al., 2012), DrugBank ( Knox et al., 2011), PubChem ( Bolton et al., 2008) or ZINC ( Irwin et al., 2012). Chemogenomic strategies have also been introduced to identify the targets of bioactive molecules in model organisms such as yeast or bacteria ( Smith et al., 2010). This has led to technological developments of large facilities enabling researchers to screen a given molecule against arrays of targets, such as kinases. In particular, it can be used (i) to predict unfavorable side effects due to off-target interactions and thus potentially decrease the attrition rate in clinical trials due to toxicity ( Kola and Landis, 2004 Lounkine et al., 2012), or (ii) to predict a new target for an approved drug and reposition it for another disease ( Ashburn and Thor, 2004 Keiser et al., 2009 Novac, 2013).Įxperimental identification of bioactive molecule targets has received much attention ( Ziegler et al., 2013). Therefore, information about the targets of bioactive molecules is crucial to understand, predict and interfere with their activity. This activity is often mediated by physical interactions with proteins or other macromolecules. This further highlights the importance of considering different classes of similarity measures between new molecules and known ligands to accurately predict their targets.Ĭontact: or information: Supplementary data are available at Bioinformatics online.Ī large number of small molecules ranging from metabolites to signaling molecules to drugs display strong bioactivity in different living systems. We also observe that different combinations of similarity measures are optimal for different molecular properties, such as the number of heavy atoms.
Protein scaffold roc curves series#
Our results indicate that the combined approach is especially efficient when no ligand with the same scaffold or from the same chemical series has yet been discovered. Results: Using a reference set of 224 412 molecules active on 1700 human proteins, we show that accurate target prediction can be achieved by combining different measures of chemical similarity based on both chemical structure and molecular shape. Computational predictions of bioactive molecule targets based on similarity with known ligands are powerful to narrow down the number of potential targets and to rationalize side effects of known molecules. In addition, the majority of bioactive molecules have more than one target, many of which are poorly characterized. However, for a significant fraction of them, the primary target remains unknown.
Motivation: Most bioactive molecules perform their action by interacting with proteins or other macromolecules.
0 kommentar(er)
