How reliable are ligand-centric methods for Target Fishing?

Abstract

Computational methods for Target Fishing (TF), also known as Target Prediction or Polypharmacology Prediction, can be used to discover new targets for small-molecule drugs. This may result in repositioning the drug in a new indication or improving our current understanding of its efficacy and side effects. While there is a substantial body of research on TF methods, there is still a need to improve their validation, which is often limited to a small part of the available targets and not easily interpretable by the user. Here we discuss how target-centric TF methods are inherently limited by the number of targets that can possibly predict (this number is by construction much larger in ligand-centric techniques). We also propose a new benchmark to validate TF methods, which is particularly suited to analyse how predictive performance varies with the query molecule. On average over approved drugs, we estimate that only five predicted targets will have to be tested to find two true targets with submicromolar potency (a strong variability in performance is however observed). In addition, we find that an approved drug has currently an average of eight known targets, which reinforces the notion that polypharmacology is a common and strong event. Furthermore, with the assistance of a control group of randomly-selected molecules, we show that the targets of approved drugs are generally harder to predict. The benchmark and a simple target prediction method to use as a performance baseline are available at

MolTarPred: A web tool for comprehensive target prediction with reliability estimation

“…However, they have the advantage of being cheap, fast and nowadays quite accurate. This has been shown in retrospective studies for a broad range of molecules (Alvarsson et al., ; Keiser et al., ; Peón et al., ) and also in prospective studies in selected molecules (Cortes Cabrera et al., ; Keiser et al., ; Mugumbate et al., ).…”

“…A particularity of this query is that vorinostat is annotated with all the 20 predicted targets (i.e., these are the 20 known targets of vorinostat within MolTarPred). This happens sometimes when the query is a clinical drug, as it tends to be more intensively studied and thus has more targets to be determined (Peón et al., ). With that said, it is not uncommon to discover previously unknown targets of clinical drugs using computational target prediction (Cortes Cabrera et al., ; Gregori‐Puigjané et al., ).…”

“…Using chemical similarity for target prediction is not without drawbacks though. This type of methods suffers from a high proportion of false negatives (true targets not predicted as such) arising from data incompleteness (Peón et al., ). On the positive side, a distinctive advantage of MolTarPred is that a target predicted with a high level of reliability is very likely to be a true target.…”

“…Indeed, the resurgence of phenotypic screening in early drug discovery (Eggert, 2013;Lee & Berg, 2013;Lee & Bogyo, 2013;Wagner, 2016) has increased the need for molecular target prediction methods. Target prediction, also known as target fishing, target deconvolution, drug-target interaction prediction or polypharmacology prediction (Byrne & Schneider, 2019;Cereto-Massagué et al, 2015;Ganesan & Barakat, 2016;Lavecchia & Cerchia, 2015;Pahikkala et al, 2015;Peón, Dang, & Ballester, 2016;Wang & Xie, 2014), aims at predicting the macromolecular targets of a given molecule (the query molecule). This problem is the reverse of screening, where the goal is to identify molecules with affinity for a given target, for example (Hoeger, Diether, Ballester, & Köhn, 2014;Naylor et al, 2009;Pereira & Williams, 2007;Xia et al, 2011).…”

Current advances in ligand‐based target prediction

“…Obviously, here comes a question with the above studies: how many ( K ) hits should be used to target prediction? To answer this question, Antonio Peón et al explored the effect of the number of hits ( K = 1, 5, 10, 15) on the recall and precision of target prediction 66 . The results show that, with the increase of K , the average recall of known targets increases gradually while the precision decreases.…”