Universal Frags2Drugs (uF2D): a proteome-wide in silico fragment-based ligand design tool
Résumé
Computational methods are extensively employed during the initial stages of drug discov-
ery process to reduce its duration and cost.(1) Fragment-based drug design (FBDD) prioritizes
binding efficiency over global binding affinity.
In our previous work, we have developed Frags2Drugs (F2D), an innovative in silico FBDD tool
aiming to design novel protein kinase inhibitors.(2) However, its applicability domain is limited
solely to protein kinases, restricting its widespread adoption. Furthermore, F2D is based on the
alignment of protein kinases, rendering it impractical for proteins belonging to distinct families.
Herein, we present Universal Frags2Drugs (uF2D), an enhanced iteration of F2D, which expands
its applicability domain to encompass all protein families. The uF2D methodology comprises two
distinct components: a fragment repositioning tool and a linker generator. The initial fragment
repositioning phase is accomplished through the adaptation and customization of a pre-existing
tool, referred to as CrystalDock.(3)
On a limited dataset consisting of 314 representative protein-ligand complexes, CrystalDock
demonstrates the ability to successfully reposition the entirety of ligand fragments in approxi-
mately 90% of the cases.
We are currently updating the fragment database with CrystalDock, and evaluating multiple
deep learning-based linker generators to incorporate into uF2D.
(1) A. V. Sadybekov, V. Katritch, Nature 2023, 616, 673–685.
(2) G. Peyrat, C. Bournez, P. Krezel, J.-M. Gally, S. Bourg, S. Aci-S`eche, P. Bonnet, Journal
of Medicinal Chemistry, submitted
(3) J. D. Durrant, A. J Keywords. Friedman, J. A. McCammon, J. Chem. Inf. Model. 2011,
51, 2573–2580.