FEP+ Pose Builder — maximizing utility and productivity in FEP simulations

The Challenge:

Bridging the Scalability Gap in FEP+

FEP+ is the industry standard for relative binding affinity prediction. However, the demanding requirement for high-fidelity initial ligand poses has traditionally imposed a significant barrier to throughput and cost-effectiveness.

Traditional FEP+ pose setup is constrained by three primary factors:

• Time and Resource Intensive: Establishing a robust protocol involves iteratively testing and comparing outputs from multiple docking and alignment algorithms (e.g., MCS alignment, Glide) to address the critical dependence of FEP+ performance on input pose. This process often consumes dedicated person days of effort.
• Steric Clashes Complication: As FEP is an all atom MD simulation based method, it can usually resolve minor protein-ligand clashes during initial equilibration. However, more severe clashes, such as when a chain threads through a ring, are not resolvable and should be avoided in the initial pose. Conventional pose generation methods are either not clash aware (e.g. ligand alignment may generate poses with perfect core alignment but with severe clashes) or too clash sensitive (e.g. docking may avoid clashes by generating ligand poses with insufficiently aligned cores resulting in suboptimal atom mapping to the FEP reference required for robust simulations).
• System Complexity: The traditional, multi-algorithmic approach necessitates managing complex external software dependencies and protocols.

FEP+ Pose Builder:

Automated, FEP-Ready Protocol Generation

FEP+ Pose Builder is a methodological advancement introduced as an integrated feature to drastically enhance accessibility, user-friendliness, and productivity within the FEP+ pipeline.

FEP+ Pose Builder transforms 2D or 3D ligand inputs into FEP-ready poses via a single, automated, four-stage protocol shown in Fig. 1:

Performance and Quantifiable Value

We benchmarked the quality of poses generated by FEP+ Pose Builder with a dataset from a recent publication¹. Shown in Fig. 2, the poses generated by FEP+ Pose Builder are similar to the poses curated manually by an expert modeler in a fraction of the time.

We also evaluated FEP+ performance of FEP+ Pose Builder poses using datasets from two publications.
Shown in Table 1, poses generated by FEP+ Pose Builder showed similar performance in FEP+ simulations
compared with the manually curated poses.

Conclusion

FEP+ Pose Builder is a critical methodological enhancement that addresses the primary bottleneck in FEP+ deployment—structural input preparation. By delivering fully automated, high-quality, and clash-aware alignment, FEP+ Pose Builder maximizes the accessibility, productivity, and cost-effectiveness of FEP+. This feature standardizes the creation of reliable FEP-ready inputs, allowing computational scientists to leverage the full power of FEP+ for molecular innovation at scale.