Plant Compounds Called Flavonoids Could Become Next-Generation Leukemia Drugs

A team of researchers spanning institutions in Nigeria, the United States, South Africa, and India argues that flavonoids, the polyphenolic compounds abundant in fruits, vegetables, soy, and tea, represent one of the most underexploited reservoirs of chemical starting material in leukemia drug discovery. In a Perspective published in Discover Chemistry, Gideon E. Mathias of the University of Calabar and colleagues set out a conceptual roadmap for how computational chemistry, rather than expensive laboratory screening, could rapidly transform these plant-derived molecules into rationally designed anti-leukemic drug candidates. Their argument arrives at a moment when the limitations of existing leukemia therapies, from toxic chemotherapy regimens to kinase inhibitors undermined by resistance, have created an urgent demand for mechanistically novel and safer treatment options.

Leukemia, a cancer of the blood-forming tissues, arises when abnormal white blood cells proliferate clonally in the bone marrow and circulation, disrupting the production of healthy red cells, platelets, and immune cells. Its four major subtypes, acute lymphoblastic leukemia, acute myeloid leukemia, chronic lymphocytic leukemia, and chronic myeloid leukemia, each carry distinct molecular drivers and therapeutic vulnerabilities. Yet despite decades of progress in understanding how these malignancies develop, clinical outcomes remain inconsistent. Chemotherapy and radiotherapy impose substantial toxicity, targeted tyrosine kinase inhibitors frequently encounter resistance and relapse, and stem-cell transplantation is constrained by donor availability, cost, and life-threatening complications. The authors contend that this therapeutic ceiling justifies a systematic search for new scaffold classes, and that flavonoids have been unjustly neglected in the hematological cancer space relative to their extensive evaluation in breast, lung, prostate, and colorectal tumors.

The appeal of flavonoids lies in their structural versatility. Isoflavones such as genistein and daidzein, and flavonols such as quercetin, have demonstrated broad anticancer activity in solid tumors by inducing apoptosis, modulating kinase signaling, regulating transcriptional networks, and sensitizing malignant cells to chemotherapy. Crucially, many of the pathways flavonoids are known to influence, including kinase signaling, epigenetic regulation, autophagy, and drug-resistance mechanisms, are precisely the pathways that drive leukemogenesis. To illustrate how these compounds might be interrogated computationally, the team anchored their analysis on daidzein, a well-characterized soy isoflavone chosen not through systematic screening but as a representative, structurally simple scaffold whose behavior could illuminate principles applicable across the entire flavonoid family.

The first layer of the analysis employed density functional theory, a quantum chemical method that maps the electronic structure of a molecule. DFT calculations revealed a HOMO energy of −6.101 electron volts and a LUMO energy of −1.703 electron volts for daidzein, yielding a HOMO–LUMO gap of 4.398 electron volts. In medicinal chemistry terms, a gap of this magnitude signals kinetic stability and reduced nonspecific reactivity, meaning the molecule is unlikely to engage in promiscuous side reactions while retaining enough electronic flexibility for selective molecular recognition. Additional descriptors, including an electrophilicity index of 3.462 electron volts, chemical hardness of 2.199 electron volts, and an electron transfer value of 0.231, reinforced the picture of a compound that balances stability with the reactivity needed to form productive interactions inside protein binding pockets.

Frontier orbital mapping and Fukui function analysis then pinpointed where those interactions would occur. The highest occupied molecular orbital distributed itself across daidzein’s aromatic framework, marking regions capable of electron donation, while the lowest unoccupied orbital localized to electron-deficient zones suited for electron acceptance. The Fukui analysis identified the oxygen atoms of hydroxyl and carbonyl groups as the primary sites for electrophilic interaction and hydrogen bonding, while the electron-rich aromatic carbons emerged as hotspots for π–π stacking with aromatic amino acid residues. Molecular electrostatic potential surfaces made the same point visually: negative potential clustered around oxygen atoms, positive potential elsewhere, providing a qualitative map that chemists could use to guide rational derivatization, strengthening hydrogen bonds, tuning electron density, or adjusting stacking potential to complement the architecture of leukemia-related protein pockets.

Those pockets were represented by two well-established drivers of the disease. FLT3, a receptor tyrosine kinase frequently mutated in acute myeloid leukemia, and LSD1, a flavin-dependent histone demethylase central to epigenetic dysregulation, were selected as conceptually relevant targets. Docking studies predicted favorable binding of daidzein to FLT3, with scores of −8.4 kilocalories per mole against the protein structure deposited as 6JQR and −7.9 kilocalories per mole against 4RT7. The more favorable 6JQR pose involved multiple hydrogen bonds anchoring the ligand in a configuration the authors interpret as consistent with longer residence time and enhanced complex stability. Notably, the interaction pattern, a hydrogen-bonding anchor to the kinase hinge region combined with aromatic stacking against gatekeeper-proximal residues, mirrors the pharmacophoric motifs of validated FLT3 inhibitors, while daidzein’s polar oxygens and planar scaffold resemble features found in known LSD1 inhibitors that occupy the FAD-adjacent pocket.

The authors are careful to frame these predictions with appropriate caution. Docking evaluates geometric compatibility and estimated binding energy but says nothing about solubility, permeability, metabolic stability, or the conformational dynamics of the protein. The study used standard rigid-receptor protocols whose scoring functions weigh shape complementarity, hydrogen bonding, hydrophobic contacts, and electrostatics, approaches useful for comparative ranking but limited in their treatment of protein flexibility and their correlation with true potency. The team also flags a subtlety specific to flavonoids: variable protonation and tautomeric states, which shift with pH and local microenvironment, can materially alter charge distribution and binding predictions. Daidzein was modeled in its neutral form as physiologically representative, with alternative ionization states documented in the supplementary material to illustrate the variability.

Drug-likeness profiling added a further dimension of realism. Daidzein satisfied all major drug-likeness rules, including Lipinski, Veber, Ghose, Egan, and Muegge criteria, with zero violations, supported by a moderate molecular weight of 254.24 grams per mole, a single rotatable bond, and an optimal topological polar surface area of 70.67 square angstroms. A consensus LogP of 2.24 indicated balanced lipophilicity, and high predicted gastrointestinal absorption pointed toward oral viability. The compound is not a P-glycoprotein substrate, reducing the risk of efflux-mediated resistance, and it produced no PAINS or Brenk alerts, with a favorable synthetic accessibility score of 2.79. Two caveats stand out: predicted inhibition of the metabolic enzymes CYP1A2, CYP2D6, and CYP3A4 raises the specter of drug–drug interactions in multi-drug leukemia regimens, and predicted blood–brain barrier penetration may be undesirable when peripheral selectivity is preferred. The authors also note that the planar, aromatic-rich scaffolds that favor π-stacking can increase the risk of binding the hERG potassium channel, a well-known contributor to cardiotoxicity, making early in silico cardiac safety screening an essential design consideration.

Looking forward, the Perspective outlines a research agenda built on several pillars. Systematic, large-scale computational screening across structurally diverse isoflavones, chalcones, and flavonols could map structure–activity trends far beyond the narrow set of scaffolds studied to date. Artificial intelligence, through generative models, reinforcement learning, and property-prediction networks, could propose flavonoid-inspired analogues optimized for potency, selectivity, and pharmacokinetics, an approach already proven in antiviral and antibiotic discovery. Molecular dynamics simulations, not performed in this study, are identified as the natural next step for validating docking poses against flexible targets like FLT3 and LSD1. Because leukemia is driven by interconnected networks rather than single targets, the authors argue that future workflows must embrace multi-target docking, network pharmacology, and systems-level modeling, potentially yielding dual-function compounds that simultaneously inhibit FLT3 signaling and modulate LSD1-mediated epigenetic states, thereby suppressing compensatory pathways and reducing the likelihood of resistance.

The final barrier is translational: flavonoids often suffer from poor aqueous solubility, rapid metabolic degradation, and limited oral bioavailability. The authors point to nanoformulation strategies using polymeric nanoparticles, liposomes, and solid lipid nanoparticles, alongside prodrug design and structural modifications such as methylation and glycosylation, as increasingly effective ways to overcome these liabilities while preserving the pharmacophoric core. Their central message is a reframing: flavonoids should no longer be regarded merely as dietary antioxidants or chemopreventive agents, but as designable chemical frameworks for structure-guided drug development. If computational predictions can be paired with rigorous biochemical assays, cellular studies, and pharmacokinetic evaluation, this largely untapped class of plant molecules may yet yield the next generation of safer, multi-target therapies for patients with leukemia.

Subject of Research: Computational evaluation of flavonoid scaffolds as potential multi-target therapeutic agents for leukemia

Article Title: Flavonoids as emerging therapeutic agents in leukemia based on conceptual frameworks computational analysis and future research direction

Article References: Mathias, G. E., Favour, E. C., Ikenyirimba, O. J., Iwuala, L. C., Ashiru, D. L., Runthala, A., Anadebe, V. C., & Ebenso, E. E. (2026). Flavonoids as emerging therapeutic agents in leukemia based on conceptual frameworks computational analysis and future research direction. Discover Chemistry, 3(1), Article 544. https://doi.org/10.1007/s44371-026-00973-2

Image Credits: AI Generated

DOI: 10.1007/s44371-026-00973-2

Keywords: flavonoids, leukemia, drug discovery, computational chemistry, density functional theory, molecular docking, daidzein, FLT3, LSD1, ADMET, natural products, medicinal chemistry