correct way to do screening

[sherifelsabbagh]

Hi

I wanted to try the framework and assess its efficacy for hit enrichment.

What I did is follows:

1. choose HDAC6 as target and download the structure PDB id: 5EEN
2. generate pharmacophore model using Pharmaconet
3. download 50 active compounds against HDAC6 from pubchem, generate decoys (50/active) using DUDE.
4. convert structures into 3D using open babel.
5. Perform screening using the generated Pharmacophore.

The results were really disappointing. Decoys get scores higher than the active and EF1% was very low.
what did I do wrong? should I generate multiple conformer for each compound ?
should I do docking and then perform screening using the model or what ?

[SeonghwanSeo]

Hello @sherifelsabbagh,

First, since PharmacoNet is a rigid-body docking approach, increasing the number of conformers improves performance. Since this program is developed for the purpose of pre-screening before docking as fine-screening, you don't have to perform docking.

Second, like many other docking programs and deep learning-based methods, the performance of PharmacoNet is target system-dependent. I recommend analyzing the correlation between the docking tool you plan to use in virtual screening against target proteins before doing a large-scale screening.

• I note that I am not familiar with the characteristics of molecules generated using DUDE.