1 of 2

Consensus scoring scheme

CODASS4: COnsensus Docking And Similarity Search

Enamine REAL

Enamine_3M

Vina-GPU+, SILCS-MC, DeepDock, DiffDock, SAIYAN*

PoseMatch

Visual analysis

Autodock-GPU

Purchase & test

39m

50k�hit�similars

Initial SBVS-based search to predict binding poses

Additional docking to complete our consensus pose prediction scheme

Our open-source pose similarity calculator

COMPRISING:

SILCS-MC

P3-Score

RFScore-VSv2

SCORCH 2.0 (inc. RFRanker)

DeepDock

TargetSF (target-specific SF trained using the 895 known ligands)

CRYSTAL LIGAND CONFORMATION (PDB 8GCY) + 895 ligands with binding poses predicted via Consensus Docking and the reliability estimator of SCORCH

2D: FP2 Tanimoto

3D: USRCAT, SILCS-Pharm, OpenFEPOPS, Autodock-SS

195m conformers

GWOVina

REAL46M_�filtered

3m

Top 5k virtual hits or “seeds”

36b compounds

50k virtual hits

Houston DR, Walkinshaw MD. “Consensus docking: improving the reliability of docking in a virtual screening context”. J Chem Inf Model. 2013

Houston DR, et al. “CODASS: A New Process for Ligand Discovery In Silico”, Conference: COMPUTATIONAL CHEMICAL BIOLOGY: probing biology with in silico tools. The University of Manchester. 2012

Our multiconformer “lead-like” subset of Enamine REAL

Maximally diverse subset of REAL46M

Fast LBVS-based search

2 of 2

*Note SAIYAN will not be used unless and until it demonstrates good performance against benchmark test datasets such as PDBbind 2021 and DUDE-Z:

http://www.pdbbind.org.cn/

https://dudez.docking.org/

Tools developed in house are shown in yellow in the previous slide

Target-specific tools trained via the addition of the 895 known ligands are underlined