Discovery of Novel Molecular Frameworks of Farnesoid X Receptor Modulators by Ensemble Machine Learning

Abstract

The bile acid activated transcription factor farnesoid X receptor (FXR) has revealed therapeutic potential as a molecular drug target for the treatment of hepatic and metabolic disorders. Despite strong efforts in FXR ligand development, the structural diversity among the known FXR modulators is limited. Only four molecular frameworks account for more than 50 % of the FXR modulators annotated in ChEMBL. Here, we leverage machine learning methods to expand the chemical space of FXR‐targeting small molecules by employing an ensemble of three complementary machine learning approaches. A counter‐propagation artificial neural network, a k‐nearest neighbor learner, and a three‐dimensional pharmacophore descriptor were combined to retrieve novel FXR ligands from a collection of more than 3 million compounds. The ensemble machine learning model identified six new FXR modulators among ten top‐ranked candidates. These active hits comprise both FXR activators and antagonists with micromolar potencies. With four novel FXR ligand scaffolds, these computationally identified bioactive compounds appreciably expand the chemical space of known FXR modulators and may serve as starting points for hit‐to‐lead expansion.

Concepts of Artificial Intelligence for Computer-Assisted Drug Discovery

“…For example, the Schneider group previously developed a SOM-based prediction of drug equivalence relationships (SPiDER) software, which tessellates a reference ligand space and predicts targets for new compounds based on assigning them to a particular winner neuron, then performing consensus scoring and statistical analysis to estimate the pseudoprobability of activity against a range of targets. Building on the success of this approach, the SOM concept was extended by the same group for many different tasks, grouping molecules with similar pharmacophoric features into clusters of functionally related compounds to generate target predictions which were then experimentally validated. − In addition, a similar technology, the counter-propagation ANN (CP-ANN), was used in a publication by Merk et al to identify novel modulators of the farnesoid X receptor, based on a combined approach using k NN and pharmacophoric screening. This study demonstrated the power of this approach for introducing useful ordering and segmentation of chemical spaces.…”

Review of in silico studies dedicated to the nuclear receptor family: Therapeutic prospects and toxicological concerns

“…Both studies present high predictive accuracy with area under the ROC curve values reaching 0.8. The third study focus on different machine learning methods (counter-propagation artificial neural network, similarity of 3D pharmacophore feature distributions method and k-nearest neighbor learner) that were optimized and combined to identify new FXR modulators molecular frameworks ( 62 ). This ensemble machine learning approach was used in a prospective screening of 3 million commercially available compounds and enable the discovery of 4 new experimentally validated FXR agonist and 2 FXR antagonist compounds with original molecular frameworks.…”

De novo Molecular Design with Generative Long Short-term Memory

“…One application of machine learning is quantitative structure-activity relationship (QSAR) modeling, which is based on the premise that the molecular structure is responsible for the molecule bioactivity and links molecular descriptors to experimentally-determined molecular properties with machine learning algorithms. [17][18][19][20][21] Compared to the 'direct' QSAR approach (i.e. inferring properties from molecular structures), 'inverse' QSAR [22,23] modeling (i.e.…”