Small-molecule inhibition of METTL3 as a strategy against myeloid leukaemia.

N6-methyladenosine (m6A) is an abundant internal RNA modification1,2 that is catalysed predominantly by the METTL3-METTL14 methyltransferase complex3,4. The m6A methyltransferase METTL3 has been linked to the initiation and maintenance of acute myeloid leukaemia (AML), but the potential of therapeutic applications targeting this enzyme remains unknown5-7. Here we present the identification and characterization of STM2457, a highly potent and selective first-in-class catalytic inhibitor of METTL3, and a crystal structure of STM2457 in complex with METTL3-METTL14. Treatment of tumours with STM2457 leads to reduced AML growth and an increase in differentiation and apoptosis. These cellular effects are accompanied by selective reduction of m6A levels on known leukaemogenic mRNAs and a decrease in their expression consistent with a translational defect. We demonstrate that pharmacological inhibition of METTL3 in vivo leads to impaired engraftment and prolonged survival in various mouse models of AML, specifically targeting key stem cell subpopulations of AML. Collectively, these results reveal the inhibition of METTL3 as a potential therapeutic strategy against AML, and provide proof of concept that the targeting of RNA-modifying enzymes represents a promising avenue for anticancer therapy.

Predicting genes associated with RNA methylation pathways using machine learning.

RNA methylation plays an important role in functional regulation of RNAs, and has thus attracted an increasing interest in biology and drug discovery. Here, we collected and collated transcriptomic, proteomic, structural and physical interaction data from the Harmonizome database, and applied supervised machine learning to predict novel genes associated with RNA methylation pathways in human. We selected five types of classifiers, which we trained and evaluated using cross-validation on multiple training sets. The best models reached 88% accuracy based on cross-validation, and an average 91% accuracy on the test set. Using protein-protein interaction data, we propose six molecular sub-networks linking model predictions to previously known RNA methylation genes, with roles in mRNA methylation, tRNA processing, rRNA processing, but also protein and chromatin modifications. Our study exemplifies how access to large omics datasets joined by machine learning methods can be used to predict gene function.

Identification and hit-to-lead exploration of a novel series of histamine H4 receptor inverse agonists.

The identification and hit-to-lead exploration of a novel, potent and selective series of histamine H(4) receptor inverse agonists is described. The initial hit, 3A (IC(50) 19 nM) was identified by means of a ligand-based virtual screening approach. Subsequent medicinal chemistry exploration yielded 18I which possessed increased potency (R-enantiomer IC(50) 1 nM) as well as enhanced microsomal stability.

Genome-wide investigation of gene-cancer associations for the prediction of novel therapeutic targets in oncology.

A major cause of failed drug discovery programs is suboptimal target selection, resulting in the development of drug candidates that are potent inhibitors, but ineffective at treating the disease. In the genomics era, the availability of large biomedical datasets with genome-wide readouts has the potential to transform target selection and validation. In this study we investigate how computational intelligence methods can be applied to predict novel therapeutic targets in oncology. We compared different machine learning classifiers applied to the task of drug target classification for nine different human cancer types. For each cancer type, a set of "known" target genes was obtained and equally-sized sets of "non-targets" were sampled multiple times from the human protein-coding genes. Models were trained on mutation, gene expression (TCGA), and gene essentiality (DepMap) data. In addition, we generated a numerical embedding of the interaction network of protein-coding genes using deep network representation learning and included the results in the modeling. We assessed feature importance using a random forests classifier and performed feature selection based on measuring permutation importance against a null distribution. Our best models achieved good generalization performance based on the AUROC metric. With the best model for each cancer type, we ran predictions on more than 15,000 protein-coding genes to identify potential novel targets. Our results indicate that this approach may be useful to inform early stages of the drug discovery pipeline.