Toward Agentic Drug Design: Experiment-Level RBFEs Across Chemical Space with NEX

SandboxAQ’s Nonequilibrium Chimeric Switching (NEX) is a new approach to relative binding free energy (RBFE) calculations designed to make high-accuracy predictions practical beyond narrow congeneric series. Where traditional RBFE methods have often required expert tuning and worked best when chemists stayed close to a single scaffold, NEX is built to support more challenging transformations — including scaffold hops, charge changes, and ring breaks — while maintaining numerical stability and consistent performance.

In this session, Mary Pitman, Lee Huntington, Dominic Rufa, and Amogh Sood will speak to how NEX combines experiment-like accuracy with a more robust and scalable workflow. Attendees will learn how the method uses a nonphysical chimeric intermediate and nonequilibrium statistical mechanics to reduce failures, lower error bars, and help teams focus less on protocol debugging and more on advancing diverse, high-value compounds. The session will also explore how NEX fits into a broader vision for automated, agentic drug discovery pipelines.

Key Learning Objectives:

• See why standard RBFE methods fail on the most valuable candidates — standard tools work well within a narrow chemical series but break down on scaffold hops, charge changes, and ring modifications. NEX is built to handle exactly these cases reliably.
• Understand what makes NEX more stable and consistent — NEX uses a novel intermediate state that reduces numerical failures and cuts the protocol debugging that consumes expert time in traditional workflows.
• Review benchmark results against real experimental data — NEX tracks closely with experimental outcomes across diverse drug targets, including the challenging scaffold hop cases where competing methods fall short.
• Learn how NEX slots into automated drug discovery pipelines — NEX integrates with enhanced sampling techniques and machine-learned force fields, supporting high-throughput workflows where teams focus on compound decisions, not computational troubleshooting.