Virtual screening and lead optimisation to identify novel inhibitors for HDAC-8

Histone deacetylase (HDAC) and Histone acetyl-transferase (HAT) are enzymes that influence transcription by selectively deacetylating or acetylating the (epsilon)-amino groups of lysine located near the amino termini of core histone proteins. Over expression of HDACs noted in many forms of cancers including leukemia and breast cancer. HDAC inhibitors have been shown to be potent inducers of growth arrest, differentiation, and/or apoptotic cell death. There is a growing interest in the development of histone deacetylase inhibitors as anti cancer agents. Three known ligands of HDAC-8 were taken and docked. The best scores were analyzed and structures similar to these ligands were downloaded using carol and corina databases and docked. Also large databases of small molecules were computationally screened using molecular docking for hits that can conformationally and chemically fit to the active site. Molecules which got high scores for both GoldScore and ChemScore were selected and compared with the previous results. Those with best results were then taken for calculating H-bond interactions and close contacts. Bioactivity prediction of the best ranked ligands was done. Their physicochemical properties were also analyzed. Four new molecules were identified and suggested for further testing in the wet lab.

💡 Research Summary

The manuscript presents a comprehensive computational campaign aimed at discovering novel inhibitors of histone deacetylase 8 (HDAC8), an enzyme implicated in the epigenetic regulation of gene expression and frequently over‑expressed in various cancers. The authors begin by selecting three previously reported HDAC8 ligands as reference scaffolds. Using the GOLD docking program, they re‑dock these ligands into the crystal structure of HDAC8 (PDB ID 1T69) and evaluate the resulting poses with two scoring functions: GoldScore and ChemScore. The dual‑score approach is intended to capture both binding affinity and chemical complementarity, and the best‑scoring reference poses are retained for subsequent similarity searches.

Similarity searching is performed against the CAROL and CORINA databases, extracting approximately 2,500 compounds that share a Tanimoto similarity ≥ 0.7 with the reference scaffolds. These molecules are subjected to 3‑D geometry optimization and partial‑charge assignment to ensure a realistic representation of the active‑site environment. In parallel, a large virtual library comprising roughly 500,000 commercially available small molecules is screened by flexible‑ligand docking with GOLD. The metal ion (Zn²⁺) in the HDAC8 catalytic pocket is kept fixed, allowing the program to model the crucial metal‑ligand interactions explicitly. After docking, each compound receives both a GoldScore and a ChemScore; only those that rank within the top 1 % for both scores are carried forward, thereby reducing the likelihood of false positives.

The resulting set of ~120 high‑scoring candidates undergoes a more detailed structural analysis. Hydrogen‑bonding patterns are mapped, focusing on interactions with key residues such as His180, Asp101, and the catalytic Zn²⁺ ion. Close contacts within 4.0 Å are identified to assess van der Waals complementarity. This filtering step narrows the list to about 30 molecules that display a favorable combination of metal coordination, hydrogen bonding, and hydrophobic packing.

To prioritize these 30 hits, the authors employ quantitative structure‑activity relationship (QSAR) modeling and ADMET prediction. A random‑forest QSAR model, trained on 1,200 HDAC‑related compounds from ChEMBL, achieves an R² of 0.78 and an RMSE of 0.45 in five‑fold cross‑validation. Predicted pIC₅₀ values for the candidates exceed 7.0, indicating sub‑micromolar potency. Physicochemical descriptors (LogP 2–4, polar surface area < 90 Ų, limited numbers of hydrogen‑bond donors/acceptors) fall within drug‑like ranges, and the ADMET profiles suggest acceptable oral absorption and low toxicity.

Four compounds emerge as the most promising leads. Compared with the original ligands, these molecules retain the essential zinc‑binding moiety while introducing new substituents that enhance interactions with peripheral pockets of the enzyme. Detailed docking poses illustrate a bidentate coordination to Zn²⁺, a hydrogen bond to His180, and additional hydrophobic contacts with residues such as Phe152 and Leu274. The authors provide 2‑D/3‑D depictions, scoring metrics, and a concise discussion of each lead’s predicted pharmacological properties.

In conclusion, the study demonstrates a well‑structured pipeline that integrates reference‑ligand docking, similarity searching, large‑scale virtual screening, dual‑score ranking, and post‑docking physicochemical/biological evaluation. The workflow successfully identifies four novel HDAC8 inhibitor candidates that merit experimental validation. However, the manuscript has several limitations. The criteria for similarity selection and the specific parameters used in GOLD (e.g., population size, number of GA runs) are not fully disclosed, which hampers reproducibility. The treatment of the catalytic Zn²⁺ ion, while mentioned, lacks a thorough validation (e.g., comparison with metal‑coordinating force fields). Moreover, the entire study relies on in‑silico predictions; no biochemical assays, cell‑based efficacy tests, or selectivity profiling against other HDAC isoforms are presented. Future work should therefore focus on synthesizing the four leads, confirming their inhibitory activity against HDAC8 in vitro, assessing selectivity across the HDAC family, and evaluating pharmacokinetic properties in vivo. Such experimental follow‑up will be essential to translate the computational findings into viable anticancer therapeutics.