Cov_FB3D: A De Novo Covalent Drug Design Protocol Integrating the BA-SAMP Strategy and Machine-Learning-Based Synthetic Tractability Evaluation

De novo drug design actively seeks to use sets
of chemical rules for the fast and efficient identification of structurally
new chemotypes with the desired set of biological properties. Fragment-based de novo design tools have been successfully applied in the
discovery of noncovalent inhibitors. Nevertheless, these tools are
rarely applied in the field of covalent inhibitor design. Herein,
we present a new protocol, called Cov_FB3D, which involves the in silico assembly of potential novel covalent inhibitors
by identifying the active fragments in the covalently binding site
of the target protein. In this protocol, we propose a BA-SAMP strategy,
which combines the noncovalent moiety score with the X-Score as the
molecular mechanism (MM) level, and the covalent candidate score with
the PM7 as the QM level. The synthetic accessibility of each suggested
compound could be further evaluated with machine-learning-based synthetic
complexity evaluation (SCScore). An in-depth test of this protocol
against the crystal structures of 15 covalent complexes consisting
of BTK inhibitors, KRAS inhibitors, EGFR inhibitors, EphB1 inhibitors,
MAGL inhibitors, and MAPK inhibitors revealed that most of these inhibitors
could be de novo reproduced from the fragments by
Cov_FB3D. The binding modes of most generated reference poses could
accurately reproduce the known binding mode of most of the reference
covalent adduct in the binding site (RMSD ≤ 2 Å). In particular,
most of these inhibitors were ranked in the top 2%, using the BA-SAMP
strategy. Notably, the novel human ALDOA inhibitor (T1) with potent inhibitory activity (0.34 ± 0.03 μM) and
greater synthetic accessibility was successfully de novo designed by this protocol. The positive results confirm the abilities
of Cov_FB3D protocol.