The DMT (Design-Make-Test) cycle drives modern drug discovery, but coordinating complex tools, manual handoffs, and disconnected workflows creates costly delays. Iambic Therapeutics has solved this challenge with an end-to-end orchestration platform that automates every stage from compound design through data analysis.
Join us for an exclusive webinar featuring:
Moderated by Nila Lê
Most drug discovery programs take years to move from a design idea to a clinical candidate. Iambic is doing it in cycles of 7 to 10 days — not by cutting corners, but by building a tightly integrated ecosystem where AI, robotics, and custom software operate as a single continuous loop. A recent webinar hosted jointly by Iambic and HighRes pulled back the curtain on exactly how that works, from molecular design to assay-ready plates to data upload.
Here's what was shared, and why it matters beyond Iambic's specific setup.
Traditional drug discovery follows a mostly linear path: identify a target, find leads, optimize for activity, refine ADME properties, and so on. At every stage, there's a real chance of hitting a wall, losing weeks, months, or in the worst case, having to start over.
Iambic's founding insight was simple: what if you could de-risk molecules before you even make them? By building AI models capable of predicting molecular properties at the design stage, they could front-load the intelligence and dramatically reduce the number of dead ends downstream.
But prediction alone isn't enough. The real competitive advantage comes from coupling those AI models with high-throughput chemistry and biology platforms and, crucially, software that keeps the entire cycle connected. The result is an iterative design-make-test loop where hundreds or thousands of compounds can be synthesized and tested in under two weeks, with every data point feeding back to retrain the models for the next round.
Two proprietary AI tools sit at the heart of Iambic's design stage.
Neuroplexer is a protein-ligand prediction model that visualizes how a molecule will interact with a target protein pocket. A calculation that would otherwise require weeks of wet lab work or expensive X-ray crystallography can be completed with a single GPU in seconds. This lets chemists evaluate binding interactions computationally before committing to synthesis.
Enchant is a multimodal transformer model trained across chemical, biological, preclinical, and clinical data modalities. By accessing and interpolating this breadth of data from the very start of the pipeline, it gives the team predictive power not just over potency, but over how a molecule is likely to behave all the way through to the clinic.
The results speak for themselves: one of Iambic's programs reached a Phase 1 clinical asset in under two years from the opening of their labs in late 2021.
The AI models and the robotics are impressive on their own. What makes Iambic's platform genuinely unusual is the in-house-built web application that connects every step: a tool they call Insight.
Senior Associate Scientist Joanna Vo described Insight as "the symphony conductor" — the software layer that drives the AI, the scientists, and the robots into coordinated motion across the full design-make-test cycle.
Insight touches every stage of the workflow:
One concrete example of Insight's impact: what used to require manually building around 30 separate files to process a set of destination plates now takes a handful of clicks and produces a clean set of five files.
Iambic's automation setup isn't a sprawling robotic floor. It's a compact, pragmatic configuration built around HighRes' FlexCart platform: a collaborative robot on a mobile cart base, surrounded by shelves holding the instruments needed for the application.
For Iambic's assay-ready plate workflow, the FlexCart is equipped with a sealer, peeler, centrifuge, solvent backfill dispenser, ambient storage, and two Echo acoustic dispensers. The small footprint matters for a lab where space is limited, and the system was designed to grow incrementally. Iambic later added a second Echo and additional storage as throughput demands increased.
The results from this modest configuration are striking. Iambic recently hit a personal record of 700 assay-ready plates in a single month — from a two-Echo system running around the clock, including overnight. As Joanna noted, "It's a very mighty system coupled with our software."
Run time optimization has been another area of focus. By reorganizing pick lists dynamically based on rack scans and implementing a serpentine dispense pattern, Iambic reduced the handling time on their Tecan liquid handler by a third and cut a large 40-plate run from four and a half hours to three and a half.
HighRes used Iambic's setup to illustrate a broader point about the current state of lab automation: the technology exists in pockets, and the gaps between those pockets are where value is lost.
When a plate washer breaks down on a work cell, for example, samples may need to be diverted to a manual benchtop process. Without an orchestration layer, that diversion means losing the automated data pipeline — fingers crossed that the right settings are used, the right plates are loaded, and the results find their way back to the data infrastructure. With orchestration, the only difference is a human loading the plates instead of a robot. The data flows continue uninterrupted.
This "automation-first" framing is central to how HighRes now positions orchestration. It's not just about large integrated systems. It's about applying the same principles (e.g., consistent protocols, automated data capture, seamless handoffs) to every part of the workflow, including the manual steps.
Iambic is currently using Cellario Scheduler™ to manage their FlexCart runs, and they access the Cellario® API via a tool called KNIME to read Echo transfer files post-run and automatically flag or recover failed transfers. A natural next step the team is exploring: a direct integration where Insight can push orders directly to Solario, eliminating the remaining manual clicks in the assay-ready plate workflow.
Several themes from Iambic's journey translate directly to labs at very different scales and stages:
Software is the multiplier. The robotics create capacity; the software determines whether that capacity translates into throughput and data quality. Iambic built Insight in-house because their needs were specific enough to justify it, but the principle, that connected software is at least as important as connected hardware, applies everywhere.
Automate the handoffs, not just the steps. The most expensive inefficiencies in Iambic's workflow weren't within any single step; they were in the coordination between teams, the file generation, the barcode tracking, and the data routing. Those are the places where Insight intervenes most decisively.
Design the data flow first. Joanna's description of Insight triggering model retraining the moment results are uploaded is only possible because the data pipeline was architected that way from the start. Treating data as a first-class outpu, not an afterthought, is what enables the closed loop.
Start where you are. Iambic didn't begin with a fully integrated facility. They started with a FlexCart, built their software incrementally, and expanded the hardware only when the workflow demanded it. The current state of their platform is the product of several years of iterative improvement, not a single upfront investment.
Iambic's design-make-test platform demonstrates that the 7–10 day drug discovery cycle isn't a theoretical future state. It's running in production today, at a company of roughly 100 people. The ingredients are AI for design, high-throughput robotics for synthesis and assay prep, and custom software that keeps every team, every instrument, and every data point in continuous communication. Getting those three things to work together is the challenge. As Iambic's journey shows, it's a solvable one.
Interested in learning more about HighRes' orchestration platform or Iambic's approach? Get in touch with us!