Design-Make-Test-Analyze (DMTA) is the iterative loop that discovers and refines therapeutic candidates. Each turn of the loop is one decision and cycle time is the currency of discovery. We have written before about why orchestration is the missing link between AI-generated experimental intent and wet-lab execution. This piece looks at the loop itself: where time is actually lost, the five concrete gaps that orchestration closes, and what that looks like in production small and large molecule workflows.
Read first: From Intent to Execution: Why Orchestration is the missing link in closed loop DMTA
Each phase of DMTA has a distinct job. Design proposes candidate molecules from prior data, models, and scientific intent. Make physically produces them: synthesis for small molecules, expression and purification for large. Test characterizes the molecules through biological and analytical assays. Analyze interprets the results, updates the model, and selects the next round of designs.
Design and Analyze are largely digital, handled by computational chemistry, modeling, and informatics platforms. Make and Test are physical: real samples, instruments, and hardware in the lab. That split creates two hand-off points where the loop is most fragile. Experimental intent has to cross from the digital world into the lab, and data has to flow cleanly back out.
Figure 1. The DMTA loop is half digital, half physical, and the joints between the two halves are where cycle time is lost.
In most labs, those hand-offs are still manual. Protocols are re-keyed and designs re-interpreted by hand as intent crosses into the lab. Samples are walked between instruments and workcells - the “sneakernet” - with high coordination overhead. Results are stitched together from files, with error-prone ID mapping and manual data merges. Scheduling conflicts across shared equipment make it difficult to scale or run multiple DMTA loops in parallel.
The pattern is consistent: cycle time is lost between the science, not in the science itself. AI now accelerates Design and Analyze, but throughput and reliability stay constrained by manual glue work in Make and Test and the hand-offs on either side of them.
Figure 2. Before orchestration: automated islands connected by manual coordination.
Cellario OS™ connects the physical lab to the digital ecosystem through an API-first product design and deep partnerships with ELN, LIMS, informatics, and AI platforms. It takes experimental intent and translates it to execution, coordinates across pools of instruments, systems, and people, and pipes data back to the analysis layer. In workflow terms, that means closing five concrete gaps, all delivered by core capabilities today:
Figure 3. Five concrete gaps, mapped onto the physical half of the DMTA loop.
Video 1. From workflow to order: a design becomes a single, structured, executable run order.
Video 2. From Order to planned action. Orders can be run immediately, on demand, or planned in to a calendar so resource allocation is clear, including pooled systems or devices.
Nanoscale synthesis shows what an orchestrated Make step looks like end to end. Building-block racks are validated, solvent added, source plates created, and reactions combined into 1536-well plates before sealing, spinning, shaking, and reacting with in-line RapidFire mass spec and LC-MS confirming that synthesis worked before compounds move downstream to functional screening on conventional HTS equipment. Orchestration removes the friction that previously gated the cycle: no manual planning bottleneck, no sneakernet labware moves, no manual data merges.
Large molecule DMTA is a different problem. Programs run fewer candidates over far longer, multi-day timelines (DNA to cells to protein) with living-cell Make steps and a complex, multi-instrument Test array spanning many workcells. Cellario OS orchestrates the full backbone: expression and purification on the Make side, then binding and functional screening plus analytical developability on the Test side, with AutoPod™ mobile robot transport between workcells, shared-instrument scheduling, and data captured back to informatics. This is not a concept: end-to-end antibody discovery labs at top 20 pharma run on this architecture today.
Figure 4. Digital representation of an installed, productive end-to-end antibody discovery lab at a top 20 pharma, powered by HighRes® lab automation.
The bottleneck is coordination, not science. With orchestration in place, the loop runs end-to-end: design flows into lab execution, execution produces contextualized data, and data flows back to update the model. Multiple campaigns run in parallel without scheduling conflicts. Every molecule and every data point carries full traceability, and scientists spend their time on science and edge cases rather than routine coordination and data plumbing.
The boundaries matter, too. Design and Analyze stay with specialist modeling and informatics platforms; orchestration connects to them; it does not replace them. That is the division of labor that makes closed-loop DMTA practical rather than aspirational.
AI reasoning, or AI Scientists, and automation hardware get the attention, but orchestration is the layer that determines whether a DMTA loop actually closes in production. The gaps are concrete: design-to-make translation, lab-wide coordination, transport, resource management, and data connectivity; and they are closable with capabilities that exist today, proven in both small molecule synthesis and multi-day biologics workflows.
Explore our applications to see how HighRes orchestration connects experimental intent to wet lab execution or schedule a demo to discuss what closed-loop DMTA can look like for your lab.