Comparative modeling and benchmarking data sets for human histone deacetylases and sirtuin families

Area of Science:

• Medicinal Chemistry
• Computational Biology
• Drug Discovery

Background:

• Histone deacetylases (HDACs) are crucial drug targets for various diseases, including cancers and neurodegenerative disorders.
• Virtual screening (VS) is an effective strategy for identifying novel and selective HDAC inhibitors (HDACIs).

Purpose of the Study:

• To construct comprehensive and maximal unbiased benchmarking data sets for HDACs, named MUBD-HDACs.
• To provide a reliable resource for validating VS approaches in HDACI drug discovery.

Main Methods:

• Developed MUBD-HDACs using methods for unbiased benchmarking, covering all four HDAC classes and 14 isoforms.
• Included 631 validated inhibitors and 24,609 unbiased decoys, ensuring chemical diversity and property matching.
• Introduced NLBScore metric to detect biases in benchmarking sets.

Main Results:

• MUBD-HDACs demonstrated maximal unbiasedness and property matching, outperforming existing benchmark sets.
• Comparative studies confirmed MUBD-HDACs' applicability to both ligand-based (LBVS) and structure-based (SBVS) virtual screening.
• The sets are chemically diverse and free from "artificial enrichment" and "analogue bias".

Conclusions:

• MUBD-HDACs represent the first comprehensive and maximal-unbiased benchmark data sets for HDACs, including Sirtuins.
• These data sets are valuable for advancing the accuracy and reliability of virtual screening in HDACI drug discovery.
• The MUBD-HDACs are publicly available for researchers worldwide.