Machine-Learning-Enabled Virtual Screening for Inhibitors of Lysine-Specific Histone Demethylase 1

Discovery of novel SOS1 inhibitors using machine learning

Overactivation of the rat sarcoma virus (RAS) signaling is responsible for 30% of all human malignancies. Son of sevenless 1 (SOS1), a crucial node in the RAS signaling pathway, could modulate RAS activation, offering a promising therapeutic strategy for RAS-driven cancers. Applying machine learning (ML)-based virtual screening (VS) on small-molecule databases, we selected a random forest (RF) regressor for its robustness and performance. Screening was performed with the L-series and EGFR-related datasets, and was extended to the Chinese National Compound Library (CNCL) with more than 1.4 million compounds. In addition to a series of documented SOS1-related molecules, we uncovered nine compounds that have an unexplored chemical framework and displayed inhibitory activity, with the most potent achieving more than 50% inhibition rate in the KRAS G12C/SOS1 PPI assay and an IC50 value in the proximity of 20 μg mL-1. Compared with the manner that known inhibitory agents bind to the target, hit compounds represented by CL01545365 occupy a unique pocket in molecular docking. An in silico drug-likeness assessment suggested that the compound has moderately favorable drug-like properties and pharmacokinetic characteristics. Altogether, our findings strongly support that, characterized by the distinctive binding modes, the recognition of novel skeletons from the carboxylic acid series could be candidates for developing promising SOS1 inhibitors.

Lysine-Specific Demethylase 1 Inhibitors: A Comprehensive Review Utilizing Computer-Aided Drug Design Technologies

Lysine-specific demethylase 1 (LSD1/KDM1A) has emerged as a promising therapeutic target for treating various cancers (such as breast cancer, liver cancer, etc.) and other diseases (blood diseases, cardiovascular diseases, etc.), owing to its observed overexpression, thereby presenting significant opportunities in drug development. Since its discovery in 2004, extensive research has been conducted on LSD1 inhibitors, with notable contributions from computational approaches. This review systematically summarizes LSD1 inhibitors investigated through computer-aided drug design (CADD) technologies since 2010, showcasing a diverse range of chemical scaffolds, including phenelzine derivatives, tranylcypromine (abbreviated as TCP or 2-PCPA) derivatives, nitrogen-containing heterocyclic (pyridine, pyrimidine, azole, thieno[3,2-b]pyrrole, indole, quinoline and benzoxazole) derivatives, natural products (including sanguinarine, phenolic compounds and resveratrol derivatives, flavonoids and other natural products) and others (including thiourea compounds, Fenoldopam and Raloxifene, (4-cyanophenyl)glycine derivatives, propargylamine and benzohydrazide derivatives and inhibitors discovered through AI techniques). Computational techniques, such as virtual screening, molecular docking and 3D-QSAR models, have played a pivotal role in elucidating the interactions between these inhibitors and LSD1. Moreover, the integration of cutting-edge technologies such as artificial intelligence holds promise in facilitating the discovery of novel LSD1 inhibitors. The comprehensive insights presented in this review aim to provide valuable information for advancing further research on LSD1 inhibitors.

Deep generative design of porous organic cages via a variational autoencoder

Porous organic cages (POCs) are a class of porous molecular materials characterised by their tunable, intrinsic porosity; this functional property makes them candidates for applications including guest storage and separation. Typically formed via dynamic covalent chemistry reactions from multifunctionalised molecular precursors, there is an enormous potential chemical space for POCs due to the fact they can be formed by combining two relatively small organic molecules, which themselves have an enormous chemical space. However, identifying suitable molecular precursors for POC formation is challenging, as POCs often lack shape persistence (the cage collapses upon solvent removal with loss of its cavity), thus losing a key functional property (porosity). Generative machine learning models have potential for targeted computational design of large functional molecular systems such as POCs. Here, we present a deep-learning-enabled generative model, Cage-VAE, for the targeted generation of shape-persistent POCs. We demonstrate the capacity of Cage-VAE to propose novel, shape-persistent POCs, via integration with multiple efficient sampling methods, including Bayesian optimisation and spherical linear interpolation.

Assisting Multitargeted Ligand Affinity Prediction of Receptor Tyrosine Kinases Associated Nonsmall Cell Lung Cancer Treatment with Multitasking Principal Neighborhood Aggregation

A multitargeted therapeutic approach with hybrid drugs is a promising strategy to enhance anticancer efficiency and overcome drug resistance in nonsmall cell lung cancer (NSCLC) treatment. Estimating affinities of small molecules against targets of interest typically proceeds as a preliminary action for recent drug discovery in the pharmaceutical industry. In this investigation, we employed machine learning models to provide a computationally affordable means for computer-aided screening to accelerate the discovery of potential drug compounds. In particular, we introduced a quantitative structure–activity-relationship (QSAR)-based multitask learning model to facilitate an in silico screening system of multitargeted drug development. Our method combines a recently developed graph-based neural network architecture, principal neighborhood aggregation (PNA), with a descriptor-based deep neural network supporting synergistic utilization of molecular graph and fingerprint features. The model was generated by more than ten-thousands affinity-reported ligands of seven crucial receptor tyrosine kinases in NSCLC from two public data sources. As a result, our multitask model demonstrated better performance than all other benchmark models, as well as achieving satisfying predictive ability regarding applicable QSAR criteria for most tasks within the model’s applicability. Since our model could potentially be a screening tool for practical use, we have provided a model implementation platform with a tutorial that is freely accessible hence, advising the first move in a long journey of cancer drug development.