Discovering highly selective and diverse PPAR-delta agonists by ligand based machine learning and structural modeling

Abstract

PPAR-δ agonists are known to enhance fatty acid metabolism, preserving glucose and physical endurance and are suggested as candidates for treating metabolic diseases. None have reached the clinic yet. Our Machine Learning algorithm called “Iterative Stochastic Elimination” (ISE) was applied to construct a ligand-based multi-filter ranking model to distinguish between confirmed PPAR-δ agonists and random molecules. Virtual screening of 1.56 million molecules by this model picked ~2500 top ranking molecules. Subsequent docking to PPAR-δ structures was mainly evaluated by geometric analysis of the docking poses rather than by energy criteria, leading to a set of 306 molecules that were sent for testing in vitro. Out of those, 13 molecules were found as potential PPAR-δ agonist leads with EC50 between 4–19 nM and 14 others with EC50 below 10 µM. Most of the nanomolar agonists were found to be highly selective for PPAR-δ and are structurally different than agonists used for model building.

Computational design of substrate selective inhibition

“…Using ISE [26][27][28][29][30], a five-fold model (see Materials and Methods) of the ncP52 set (31 fragments) vs. the random set (7104) was constructed. The AUC was 0.97 (see ROC curve in S4A Fig) [31].…”

“…We may tackle that question by using an example from "real world" screening. In HTS, it is expected (but not guaranteed) to find about 1 active molecule out of a thousand screened [23,30]. Therefore, to mimic that level of success in VS (Virtual Screening), a model of actives vs. assumed inactives should also be based on a 1 (true positive):1000 (true negatives) ratio.…”

DeepSnap-Deep Learning Approach Predicts Progesterone Receptor Antagonist Activity With High Performance

“…In silico computational approaches such as machine learning (ML) methods are useful tools for discovery agonists and antagonists, particularly in modeling of ligand-binding protein activation with an increasing number of new chemical compounds synthesized (Banerjee et al, 2016;Niu et al, 2016;Asako and Uesawa, 2017;Wink et al, 2018;Bitencourt-Ferreira and de Azevedo, 2019;Da'adoosh et al, 2019;Kim G. B. et al, 2019). Among in silico approaches, both qualitative classification and quantitative prediction models by quantitative structureactivity relationship (QSAR) methods were reported using a large collection of environmental chemicals (Zang et al, 2013;Niu et al, 2016;Norinder and Boyer, 2016;Cotterill et al, 2019;Dreier et al, 2019;Heo et al, 2019).…”

Prediction of novel mouse TLR9 agonists using a random forest approach

“…We hypothesized that modern in silico high-throughput screening (HTS) methods may improve the ability to identify novel highly active TLR9 ligands. In silico screening, also known as virtual screening (VS), has been widely used to enrich datasets with compounds that have a higher probability of binding to the target of interest [3][4][5], and has an advantage over traditional screening or physical HTS due to its massively parallel processing ability; hence millions of compounds can be assessed economically in parallel. This is particularly important when the search space for potential ODNs TLR9 ligands is taken into consideration.…”

“…ML is an umbrella term for statistical models built to discover patterns in existing data to explain unseen data. ML models are very powerful tools that have been used in the past to predict and classify the pharmacokinetics or toxicological profiles of compounds [12], predict biological activities or toxicity [13] and assist in screening and optimization of compounds [5].…”