Computational fragment-based screening using RosettaLigand: the SAMPL3 challenge.

Area of Science:

• Computational chemistry
• Drug discovery
• Structural biology

Background:

• The SAMPL3 (Statistical Assessment of Molecular Sampling and Physical Models) challenge offers a blind testing environment for virtual screening methods.
• Evaluating virtual fragment screening protocols is essential for advancing drug discovery techniques.

Purpose of the Study:

• To assess a virtual fragment screening protocol using RosettaLigand against bovine pancreatic trypsin within the SAMPL3 challenge.
• To identify factors influencing the success of fragment-based virtual screening.

Main Methods:

• Screening of a 500-fragment Maybridge library using RosettaLigand.
• Evaluation of virtual screening performance in a blind testing scenario.
• Analysis of the impact of known active compounds, parameter selection, and receptor ensembles.

Main Results:

• Virtual screening performance is enhanced by the availability of known active compounds and optimized parameters.
• Accurate ligand orientation and conformation depend on appropriate partial charge calculation methods.
• Using multiple receptor ensembles does not consistently improve docking enrichment compared to individual receptors.

Conclusions:

• Blind testing environments like SAMPL3 are critical for validating virtual screening approaches.
• Fragment screening success can be significantly increased through careful selection of receptor structures, protein flexibility, conformational sampling, and accurate partial charge assignments.