Discovery of novel compounds for cancer therapy via targeting cancer stem cells and lysine-specific histone demethylase-1

Abstract

Cancer remains a major global health challenge due to uncontrolled proliferation, metastasis, and therapeutic resistance, underscoring the need for more effective and selective treatment strategies. This thesis integrates synthetic chemistry, machine learning-assisted virtual screening, and multiscale computational modelling to develop two classes of targeted anticancer agents: pleuromutilin-derived inhibitors targeting cancer stem cell (CSC)-associated pathways and novel small-molecule inhibitors of lysine-specific demethylase-1 (LSD1), a key epigenetic regulator implicated in tumour progression.

In the first part of this work, a structure-guided framework was established for the design and synthesis of pleuromutilin-based anticancer agents. By analysing electrophilic reactivity, bioelectronic balance, and steric constraints within the pleuromutilin scaffold, selective esterification at the R₂ position was identified as an effective strategy to enhance biological potency while preserving core pharmacophoric features. Systematic structure–activity relationship studies demonstrated that para- and meta-substituted electron-withdrawing benzoates, particularly fluorinated derivatives, significantly improved antiproliferative and anti-CSC activities. Lead compounds WYA015 and WYA087 exhibited potent efficacy, highlighting the value of metabolically stable fluorinated motifs for enhancing intracellular persistence and target engagement.

The second major component of this thesis focuses on the discovery of potent LSD1 inhibitors. It has significant importance in cancer therapy, as LSD1 overexpression contributes to tumour progression and malignancy. This study employed an integrated strategy combining machine-learning-based virtual screening, bioassay validation, and ligand–target interaction analysis. Virtual screening identified 29 candidate compounds, of which 17 demonstrated micromolar-level enzymatic inhibition in experimental assays. Among them, compound L01 showed strong potential as an LSD1 inhibitor, exhibiting antileukemic activity (IC₅₀ = 24 μM) and confirmed binding to LSD1 in Kasumi-1 cells. Microscale thermophoresis (MST) and surface plasmon resonance (SPR) analysis confirmed the direct binding between L01 and LSD1 inferred from our in vitro studies. Molecular docking and molecular dynamics (MD) simulations further supported the interaction, revealing a binding affinity of L01 comparable to that of the well-characterised reversible inhibitor SP2577. Molecular docking and MD simulations further demonstrated that ligand binding to LSD1 is governed not by isolated strong interactions but by a delicate balance between favourable hydrophobic and van der Waals (VDWAALS) contacts and unfavourable electrostatic or solvation effects. Key residues, including ALA160, THR444, VAL622, and GLU612, were identified as critical binding hotspots, whereas LYS481 and arginine-rich regions contribute to detrimental interactions. Furthermore, the roles of hydrogen bonding and electrostatics were found to be context-dependent, modulating affinity without dominating the binding landscape. Notably, ligand L01 adopts a stable conformation within the catalytic pocket, maintaining persistent interactions with anchors such as ALA160. These mechanistic insights into the energetic and structural drivers of LSD1–ligand recognition provide a rational framework for the design of novel inhibitors that optimise non-polar contacts while mitigating unfavourable electrostatic penalties. Absorption, distribution, metabolism, excretion, and toxicity (ADMET) analysis revealed that L01 possesses excellent Caco-2 permeability and high target specificity; however, its moderate drug-likeness and potential toxicity issues underscore the need for further optimisation. Through computational modelling and experimental validation, our study identifies L01 as a promising LSD1 inhibitor with strong potential for optimisation to improve bioavailability and safety in cancer therapy.

Building on this foundation, we conducted an in-depth mechanistic analysis of L01 analogues by integrating MD simulations with energy decompositions using molecular mechanics/Generalised Born surface area (MM/GBSA) and molecular mechanics/Poisson-Boltzmann surface area (MM/PBSA). Several analogues exhibited substantially improved potency, demonstrating the suitability of the L01 scaffold for structure-based optimisation. Stable binding poses were characterised by low ligand and backbone root mean square deviation (RMSD) values and minimal fluctuations in key recognition residues. The hydrogen-bond and energy decomposition results indicate that ligand potency is governed not simply by the number of persistent contacts, but by the nature and location of those interactions. Increased occupancy of hydrogen bonds involving ARG145, and to a lesser extent SER115, GLU612, THR444, and SER580, tends to associate with higher IC₅₀ values, suggesting these contacts may reflect less productive binding modes. In contrast, interaction with LYS481 may support improved activity, which is opposite to the former study, indicating a context-dependent role of this residue. Energetically, TYR581, TRP571, ALA577, and ALA147 act as major stabilising residues, whereas ARG631, ARG141, ASN617, ARG124, and THR386 contribute unfavourably, highlighting key regions for future ligand optimisation efforts. Guided by these mechanistic insights, a L01 analogue library had been designed. Through Schemes 2 and 3, L01 and thirteen L01 analogues had been synthesised. Overall, this work establishes L01 as a viable lead compound and provides a quantitative, mechanism-informed framework for designing next-generation LSD1 inhibitors.
Together, these studies advance the mechanistic understanding of both pleuromutilin-derived anti-CSCs agents and LSD1 inhibitors, providing a robust blueprint for structure-based optimisation. The integrated approach demonstrated in this thesis—combining synthetic innovation, machine learning, and rigorous computational and experimental validation—offers a powerful platform for the discovery of next-generation, high-selectivity anticancer therapeutics.

Date of Award	18 Jul 2026
Original language	English
Awarding Institution	University of Nottingham
Supervisor	Bencan Tang (Supervisor), Jonathan D. Hirst (Supervisor) & Yongmei Cai (Supervisor)