What features could Schrödinger add to its FEP+ software to increase accuracy in predicting protein-ligand binding affinities?

Introduction

To improve Schrödinger's FEP+ software for increased accuracy in predicting protein-ligand binding affinities, we need to carefully analyze the current product, user needs, and potential areas for enhancement. I'll approach this challenge by examining key stakeholders, identifying pain points, generating innovative solutions, and proposing a strategic implementation plan.

Step 1

Clarifying Questions (5 mins)

• Looking at the product context, I'm thinking FEP+ is primarily used by computational chemists and drug designers. Could you confirm the primary user base and their key use cases?

Why it matters: Determines the focus of our improvements and ensures we're addressing the right user needs. Expected answer: Primarily used by pharmaceutical researchers and computational chemists for drug discovery. Impact on approach: Would tailor features to scientific workflow and accuracy requirements.

• Considering the emphasis on accuracy, I'm curious about the current performance benchmarks. What's the current accuracy level of FEP+ compared to experimental data, and how does it stack up against competitors?

Why it matters: Helps identify the gap we need to close and prioritize improvements. Expected answer: FEP+ is among the top performers but still has a mean absolute error of ~1 kcal/mol. Impact on approach: Would focus on reducing this error and potentially incorporating new computational methods.

• Given the complexity of protein-ligand interactions, I'm wondering about the computational resources required. What are the current limitations in terms of system size or simulation time that users face?

Why it matters: Identifies potential bottlenecks and areas for optimization. Expected answer: Limited to systems of ~50,000 atoms and simulation times of ~100 ns. Impact on approach: Would explore ways to enhance performance or implement more efficient algorithms.

• Thinking about the broader ecosystem, how does FEP+ currently integrate with other tools in the drug discovery pipeline?

Why it matters: Ensures our improvements align with the overall workflow and don't create friction. Expected answer: Integrates with molecular docking tools and ADME prediction software. Impact on approach: Would consider enhancing existing integrations or adding new ones to streamline the workflow.