At SapioCon 2026, Rob Brown’s ecosystem session captured something that has been building across the biopharma industry for some time. The conversation around AI in drug discovery has matured. There is less focus on whether AI belongs in the lab and more on the harder, practical question: how do you make it work inside real scientific workflows, with tools organizations already trust, without compromising data governance or scientific rigor?

The “Human Middleware” bottleneck

The challenge is one many scientists will recognize. Organizations have spent years building and validating sophisticated computational methods and applications: molecular docking platforms, ADMET prediction engines, and retrosynthesis algorithms. Scientists trust these tools not simply because they produce consistent outputs, but because extensive scientific research and years of validation by computational experts underpin them.

The problem is access. These capabilities largely operate in isolation, and getting the most from them often means either navigating multiple disconnected systems or waiting on a computational specialist who has the skills and access to run the analysis. In either case, critical scientific reasoning ends up scattered across platforms and inboxes rather than captured where it belongs. inside the experiment record.

As Brown explained, Sapio’s response is not to replace these tools but to connect them directly into Sapio Elain. As a result, every analysis, result, and decision is captured in a single governed record, available across the organization in real time.

Four partners, one workflow

The centerpiece of the session was a live illustrative medicinal chemistry workflow, with partners from Cadence Molecular Sciences (OpenEye), Simulations Plus, CCDC and Elsevier each presenting their integration on stage, supported by NVIDIA AI infrastructure. The scenario is one that any medicinal chemist can relate to. Starting from a known competitor compound, the goal is to identify novel candidates likely to be active against the same target but outside patent-protected chemical space and assess which can realistically be progressed.

Each step was executed through natural language prompts inside Sapio Elain, with results returned directly into the experiment record. No exports, no platform switching, no reconstructing context after the fact.

Cadence Molecular Sciences (OpenEye): Escaping the Patent Trap

  • Using OpenEye’s Orion platform, Elain runs a 3D shape similarity search against the Enamine HTS library. Rather than relying on 2D structure, the search identifies molecules with a similar 3D shape to the competitor compound. These candidates look structurally different and sit outside patent space but may interact with the target in a comparable way. The search runs asynchronously and returns results directly into the experiment record.

Simulations Plus: Developability First

  • With a large set of candidates returned, ElainN calculates ADMET profiles using Simulations Plus, selecting metabolism and toxicity panels to filter the dataset down to a more tractable set. Predicted properties are returned inline, allowing the scientist to prioritize without leaving the notebook.

Sapio Registry: The Governance Glue

  • The workflow stays anchored to internal standards through the Sapio Registry. Prepared protein targets and proprietary compound records are retrieved directly, ensuring external computational exploration is always governed by internal scientific truth.

CCDC: The Gold Standard for Docking

  • Elain docks the filtered candidates into the protein using CCDC Gold. Scores and 3D binding poses are returned inline within the experiment, allowing the scientist to assess how each candidate engages the active site without leaving the record.

Elsevier: Makeability as a Constraint

  • The final step illustrates a capability that becomes critical when candidates need to be synthesized rather than purchased. Using Reaxys data curated from 16 million organic reactions, Elain calculates synthetic accessibility scores across the candidate set. Where synthesis is required, a retrosynthesis analysis proposes viable routes, each with a confidence score. Elain then builds the full med-chem experiment directly in the notebook.

The scientist stays in control

What this workflow illustrates is a deliberate position on how AI should operate in scientific research and, equally, what it should not do.

The Elain ecosystem does not override the tool choices organizations have already made. Instead, it surfaces whichever platforms have been approved and validated, configured by the teams best placed to make those decisions. The scientist determines what question to ask, reviews the results, decides which candidates to advance, and chooses the next step. AI handles coordination and capture, while scientific judgment stays with the scientist. That distinction matters for organizations that need AI to accelerate discovery without the unpredictability of fully autonomous systems.

It is also worth being precise about what kind of AI is doing what. Orion, Gold, and Reaxys are not AI systems; they are validated methods and applications scientists have relied on for years, with repeatable outputs built on deep scientific foundations. Sapio ELaiN uses AI to interpret natural language, coordinate access to those tools, and capture every result in context. What changes is how scientists reach them and the fact that every result ends up in the record rather than outside it.

The Sapio Elain ecosystem

The workflow demonstrated at SapioCon is one illustration of what that looks like in practice. The broader ecosystem spans molecular modeling and structure-based design, predictive analytics, cheminformatics, bioinformatics and biological insight. Full partner details are available in the ELaiN ecosystem announcement and on the Sapio ELaiN ecosystem page.

Key Points

  • Biopharma and biotech organizations rely on validated computational methods and applications that largely operate in isolation, creating access barriers, context loss and fragmented experimental records.
  • Sapio Elain, Sapio’s AI Lab Notebook, connects those tools into the workflow through natural language prompts, with every result captured in a single governed experiment record.
  • A live workflow at SapioCon 2026 illustrated how OpenEye, Simulations Plus, CCDC Gold and Elsevier Reaxys can be coordinated within a single ELaiN session, from competitor compound to synthesizable candidate.
  • The Elain ecosystem surfaces approved tools and keeps the scientist in control at every step. It does not replace validated methods or override organizational choices about which tools to use.
  • Sapio Elain uses AI to coordinate access to established scientific methods, ensuring research stays grounded in trusted, validated outputs while removing the friction of fragmented workflows.