مقالات پذیرفته شده در نهمین کنگره بین المللی زیست پزشکی
AI-Guided Design, Docking, and ADMET Profiling of Novel HDAC2 Inhibitors for Cutaneous T-Cell Lymphoma Therapy
AI-Guided Design, Docking, and ADMET Profiling of Novel HDAC2 Inhibitors for Cutaneous T-Cell Lymphoma Therapy
Elina Khanehzar,1Fatemeh Shams,2Amir Sajad Jafari,3,*
1. Student Research Committee, Shiraz University of Medical Sciences, Shiraz, Iran / Medicinal and Natural Products Chemistry Research Center, Shiraz University of Medical Sciences, Shiraz, Iran 2. Student Research Committee, Shiraz University of Medical Sciences, Shiraz, Iran / Medicinal and Natural Products Chemistry Research Center, Shiraz University of Medical Sciences, Shiraz, Iran 3. Department of Basic Sciences, School of Veterinary Medicine, Shiraz University, Shiraz, Iran / Medicinal and Natural Products Chemistry Research Center, Shiraz University of Medical Sciences, Shiraz, Iran
Introduction: Cutaneous T-cell lymphoma (CTCL) is a rare but challenging non-Hodgkin lymphoma of skin-homing T lymphocytes. Despite therapeutic advances, many patients relapse or progress, underscoring the need for novel therapeutic options. Epigenetic dysregulation is central to CTCL biology, with histone deacetylase 2 (HDAC2) frequently reported as upregulated in malignant T cells. High HDAC2 expression correlates with disease progression and poor prognosis in T-cell lymphomas and other cancers (CRC, breast, GBM).Vorinostat (SAHA) is FDA-approved as an HDAC inhibitor for CTCL, but its clinical utility is limited by systemic toxicity, modest selectivity, and incomplete responses. Artificial intelligence (AI) provides a powerful opportunity to accelerate drug discovery by generating scaffold libraries guided by pharmacophore knowledge and filtered through drug-likeness criteria. This study applied AI-driven scaffold generation, docking, and ADMET profiling to identify novel HDAC2 inhibitors with potential therapeutic relevance for CTCL.
Methods: A library of 100 candidate scaffolds was generated using a meta-prompt strategy. The AI was instructed to derive pharmacophore motifs from HDAC inhibitors in clinical use or trials, incorporate diverse zinc-binding groups (hydroxamic acids, benzamides, ortho-anilides, thiols), explore heteroaromatic scaffolds (indoles, quinazolines, pyrimidines) for pocket occupancy, and embed drug-likeness constraints (LogP 2–4, TPSA <120 Ų, absence of toxicophores). Docking simulations were conducted against HDAC2 (PDB ID: 7MOZ, chain A) using Glide. The top 10% of molecules were shortlisted, and 10 final candidates were advanced for ADMET profiling. Predictions included lipophilicity, TPSA, QED, solubility, gastrointestinal absorption, cytochrome P450 inhibition, hERG channel liability, and Ames mutagenicity. Vorinostat was used as the positive control.
Results: Docking scores for the 10 shortlisted promising candidates ranged from –9.064 to –4.611 kcal/mol. The top three were Candidate No. 1 (–9.064 kcal/mol), Candidate No. 2 (–8.276 kcal/mol), and Candidate No. 8 (–8.064 kcal/mol). Additional candidates scored between –7.691 and –7.456 kcal/mol, while the weakest values were –6.452, –5.069, and –4.611 kcal/mol. For comparison, the positive control Vorinostat scored –5.392 kcal/mol against HDAC2. Thus, all but one candidate exhibited stronger binding affinity than the approved reference drug.
ADMET profiling highlighted favorable physicochemical and pharmacological properties. Candidate No. 1 displayed TPSA 78.43 Ų and LogP 2.1, with high solubility and gastrointestinal absorption. Candidate No. 2 exhibited TPSA 105.73 Ų and LogP 1.36, alongside good solubility and absorption. Candidate No. 8 had TPSA 75.36 Ų and LogP 1.3, with excellent solubility and high absorption. Importantly, all three showed absence of CYP inhibition, no hERG liability, no mutagenicity, and no PAINS alerts.
Conclusion: This AI-guided drug discovery workflow successfully identified 10 prioritized HDAC2 inhibitor candidates for CTCL from an initial library of 100 molecules. Nearly all AI-designed candidates demonstrated stronger docking affinity than Vorinostat, with Candidate Nos. 1, 2, and 8 achieving the most favorable profiles. These leading molecules combined robust binding to HDAC2 with consistently safe and drug-like ADMET properties. The results underscore the potential of AI-driven design, docking, and ADMET profiling to deliver next-generation HDAC2 inhibitors with therapeutic promise in cutaneous T-cell lymphoma.