Revilico engine benchmarks against experimental data - Revilico

What this tab covers

This tab reports how Revilico’s engines perform on public benchmark data sets, compared against experimental measurements. Each benchmark page states:

which data sets were used;
which settings were run;
how the numbers were calculated;
where the results have known limits.

Every number labelled as a Revilico engine result comes from our own runs with the engine’s default settings. Nothing was tuned per target. Reference results from other methods are labelled as published data from their authors.

Available benchmarks

RevFEP (free energy)

Relative binding free energies on the public JACS and Merck protein–ligand sets: 17 targets and 389 ligands, compared with experiment and with Schrödinger’s published FEP+ results.

RevFEP at a glance

Mean per-target RMSE and Spearman correlation for RevFEP and FEP+ — Averages of the per-target statistics on the JACS and Merck sets. RevFEP numbers are our runs; FEP+ numbers are Schrödinger's published results (release 21-4). See the RevFEP benchmark page for per-target results, an interactive explorer and the methods.

Set	Targets	RevFEP RMSE	FEP+ RMSE	RevFEP ρ	FEP+ ρ
Overall	17	1.28	0.79	0.48	0.68
JACS	6	1.00	0.68	0.68	0.77
Merck	11	1.43	0.85	0.38	0.64

RMSE is in kcal/mol. ρ is the Spearman rank correlation.