FEP+ Protocol Builder
ML-powered optimization of FEP+ models for challenging targets
FEP+ Protocol Builder is an automated machine learning workflow for FEP+ model optimization. It saves scientists time and improves the chances of successfully enabling discovery with FEP+ by identifying an optimized predictive model for your target.
Improve FEP+ accuracy with physics-driven machine learning for the most challenging targets to confidently use FEP+ prospectively in your program
Reduce compute cost across the lifetime of your project with automated cost optimization of FEP+ models
Generate optimized FEP+ models 4x faster than manual optimization, saving weeks of researcher time across diverse disease areas through efficient exploration of parameter space
Expedite FEP+ use for challenging systems with a fully automated workflow
Fully automated FEP+ Protocol Builder generated optimized FEP+ protocol with improved pairwise RMSE compared to default model. All protocols were performed with the same 15 congeneric ligands with modifications at 2 R-groups with known affinity data from PDB ID: 2IVV. FEP+ Protocol Builder was applied to quickly and rigorously validate the amenability of the target to the Schrödinger platform.
| Workflow | Description | Library Size | Time to screen 4.0B (days) | Time to screen 6.5B (days) | Storage space for 6.5B (TB) |
|---|---|---|---|---|---|
| Quick Shape | Combination of 1D-SIM* prefilter and Shape CPU Screening | > 4.0 billion | 5.2 | 5.5‡ | 0.4 |
| Shape GPU | GPU-accelerated 3D screening | < 5.0 billion | 4.6 | 7.5‡ | 33 |
| Shape CPU | CPU-based 3D screening | < 10 million | NA | NA | NA |
FEP+ Protocol Builder models routinely outperform those from human experts across diverse protein targets
In a recent study, FEP+ Protocol Builder outperformed human experts in producing a predictive model across ten diverse targets where default FEP+ settings did not produce an appropriately accurate protocol (RMSE > 2.5 kcal/mol) and were able to generate FEP+ models for systems where experts failed.
FEP+ Protocol Builder was run in a completely automated fashion following a rigorous training/test set split. On average, the reduction in turnaround time for the final optimization model went from 27 days to 7 days, a 4x acceleration, saving on average 20 days per project.
| Disease area | Target class | Target | Expert Protocol RMSE (kcal/mol) | Expert Protocol RMSE (kcal/mol) |
|---|---|---|---|---|
| Oncology | Bcl-2 | MCL1 | 1.5 |
1.1 |
| Neurology | ATPase | P97 | 1.3 |
1.0 |
| Oncology | Nuclear receptor | ESR1 | 3.1 |
2.0 |
| Pain, addiction, oncology | GPCR | mOR | 2.4 |
2.2 |
| Pain, addiction | GPCR | dOR | 2.2 |
1.3 |
| Hematology, oncology | ADP-ribosyltransferase | TNKS2 | 2.2 |
1.1 |
| Pain, addiction, neurology | GPCR | KOR | 2.1 |
1.7 |
| Renal | Aspartic protease | Renin | 1.8 |
1.6 |
| Oncology and rheumatology | Cysteine protease | MALT1 | 2.5 |
1.5 |
| Oncology | Receptor tyrosine kinase | RET | 1.9 |
0.8 |
Webinar
Expediting FEP+ model optimization for challenging systems with a fully automated, machine learning-driven workflow
FEP+ is a powerful predictive technology in drug discovery – with applications from hit discovery through lead optimization. A critical first step in deploying FEP+ is to validate and optimize the model for the protein-ligand system of interest.
Watch WebinarMachine learning driven FEP+ Protocol Builder workflow
Recommended inputs
- One experimentally resolved protein-ligand structure or a computationally generated protein-ligand binding mode hypothesis
- 10 or more congeneric ligands, ideally 20 for better statistics, are needed. The ligands should have known affinity data spanning at least two to three orders of magnitude
- FEP+ Protocol Builder is bundled with FEP+ and requires a minimum of 20 licenses for optimal use
Available as software or a service
- Leverage your internal resources or work with Schrödinger’s team of experts and large-scale computational resources to optimize your next FEP+ model.
Publications
-
FEP+ Protocol Builder: Optimization of free energy perturbation protocols using active learning
de Oliveira C, et al. J. Chem. Inf. Model. 2023, 63, 17, 5592–5603.
-
Advancing drug discovery through enhanced free energy calculations
Abel R, et al. Acc. Chem. Res. 2017, 50, 7, 1625–1632.
-
Accurate and reliable prediction of relative ligand binding potency in prospective drug discovery by way of a modern free-energy calculation protocol and force field
Abel R, et al. J. Am. Chem. Soc. 2015, 137, 7, 2695–2703.