Artificial intelligence is transforming drug discovery. From molecular design and synthesis planning to data analysis and decision-making, AI promises to dramatically increase the speed at which new therapeutic ideas can be generated and evaluated.
But faster design is only valuable if the rest of the discovery workflow can keep pace.
As AI expands our ability to explore chemical space, the number of compounds entering early discovery workflows is set to increase significantly. For many organisations, generating candidate molecules may soon become easier than making, testing and learning from them. In other words, the limiting factor in drug discovery is shifting away from idea generation and towards the productivity of the DMTA cycle itself.
This matters because the commercial pressures on drug discovery continue to intensify. The industry's focus has shifted toward increasingly complex biological targets, while advances in patient stratification are driving demand for more personalised therapies. Together, these trends require organisations to investigate larger numbers of compounds, generate more data and make better decisions – all while controlling costs and reducing development timelines.
For decades, improvements in drug discovery productivity have been delivered through advances in individual parts of the workflow, from high-throughput screening and laboratory automation to computational modelling. However, incremental gains within isolated functions are unlikely to be sufficient for the next era of drug discovery. As the volume of compounds and data continues to grow, bottlenecks that were once manageable risk becoming major constraints on innovation.
The future will belong to organisations that stop viewing Design, Make, Test and Analyse as separate activities and instead optimise DMTA as a single integrated system. Achieving the iteration velocity needed for AI-enabled drug discovery will require more than faster algorithms. It will demand new approaches to data infrastructure, decision-making, synthesis, purification, testing and the interfaces that connect them.
Why productivity is the next big challenge in drug discovery
Despite decades of innovation in areas such as high-throughput screening, laboratory automation and molecular modelling, improvements in drug discovery productivity have been difficult to achieve. Many of the easier targets have already been addressed, while the rise of personalised medicine is increasing the pressure to generate more data and evaluate more compounds efficiently.
Against this backdrop, AI offers enormous potential. However, its success will depend not only on generating better ideas, but on increasing the speed at which those ideas can be translated into experimental learning.
One consequence of this AI-driven revolution is that it will increase the number of compounds passing through early DMTA cycles. This will demand a dramatic increase in the ‘iteration velocity’ of the whole DMTA process – something that isn’t going to be achieved by making incremental adjustments to disconnected processes within departmental silos, as we’ve done in the past. Instead, we need to take a more holistic view of DMTA, one in which the focus is on exploring molecular space efficiently and producing high-quality data that enables better decisions to be made more quickly.
Looking across the DMTA workflow, three areas stand out as particularly important opportunities for improvement. The first is purification, where labour-intensive handling steps continue to constrain throughput despite advances in synthetic chemistry. The second is testing, where emerging biological models such as organoids promise richer and more predictive data but remain difficult to deploy at scale. The third is the interface between Make and Test, where traditional workflows still impose delays and barriers between chemistry and biology.
Together, these challenges are likely to become the defining constraints on DMTA iteration velocity over the coming decade.
The industry is already beginning to experience this shift. Investments in AI are accelerating the front end of drug discovery, enabling teams to explore larger areas of chemical space and generate promising candidates more efficiently than ever before. However, increasing design capacity creates limited value unless the downstream discovery workflow can keep pace. As a result, attention is increasingly turning to the productivity of the entire DMTA system, and to the bottlenecks that limit its iteration velocity.
Overcoming structural inertia in pharma
What would an optimum DMTA cycle look like if we started from a blank sheet of paper?
First of all, you’d need to overcome the structural inertia in large areas of pharma, and put in place more efficient IT and governance structures. Key to this will be using a unified lab IT platform, acquiring highly granular, time-stamped data on key performance indicators that is free from temporal gaps. Such an interface could then be used to track project progression, with the data delivered by it being used to ensure that processes are running at maximum efficiency, and that bottlenecks are identified and resolved on a continuous cycle.
Related to this is the need to make more efficient decisions on projects, by placing more emphasis on data-driven decision frameworks that help reduce subjective bias and improve consistency in project decision-making. This would enable samples to be rapidly funneled into the right assay without humans necessarily being in the loop, and resolve the commonplace issue of projects not being killed early enough because the staff are deeply invested in success, and hence reluctant to admit defeat (and even learn from such failures).
With both these changes implemented, it’s much more likely that companies will be able to reap the full benefits of rethinking the technologies used in DMTA processes – and I discuss some ways of doing this below.
Bringing synthesis and purification up to speed
When it comes to streamlining DMTA, Test is already highly automated, thanks to the learnings transferred from high-throughput screening in the 1990s. And with advances in testing technologies such as phenotypic assays, deep metabolomic analysis and increasingly sophisticated organoid models, there is every reason to believe that we will soon be able to generate richer and more predictive biological data at scale. These next-generation testing systems have the potential to improve decision quality dramatically, provided they can be manufactured, standardised and deployed with sufficient throughput to support future DMTA workflows.
That’s fine as far as it goes, but improvements in Test have historically had limited impact on overall discovery timelines because many of the obvious workflow and organisational efficiencies have already been realised. The bottlenecks today are often in the Make phase where we still rely on the expertise of chemists to guide reactions to a successful outcome.
So if DMTA is going to be transformed as it needs to be, we need to find ways of ‘moving the needle’ on the traditional throughput limitations of synthetic chemistry. One appealing idea being studied is using machine learning to automatically generate new synthetic protocols from a standardised literature database. This can then be coupled with robotic automation, which has the benefit of enabling existing glassware and synthetic processes to be used without major changes.
Better still, we could turn to ‘high-throughput experimentation’ for synthetic chemistry. Microscale chemistry where reactions are carried out in wells on a microtiter plate, aided by high-precision tools such as acoustic dispensing can allow us to explore chemistry space efficiently. And although microtiter plates are associated with biochemical reactions, with material-science expertise and some ‘out-of-the-box’ thinking, there’s no reason why they could be designed to be compatible with commonly used reactants and organic solvents.
The appeal of this approach is reflected in growing industry adoption of high-throughput experimentation (HTE) platforms, which enable hundreds or even thousands of reactions to be screened in parallel. By generating richer reaction data and exploring chemical space more efficiently, HTE is increasingly being viewed as a foundational capability for future DMTA workflows.
We could also look to overcome the purification bottleneck by re-engineering the manual operations between synthesis and chromatography. Or we could eliminate the need for purification entirely by choosing ‘clean’ single-pot multi-component reactions or using solid-phase technology to enable direct to biology testing.
Streamlining the Make/Test interface
In a way, part of the problem with DMTA is the abbreviation itself, because it embeds the thinking that Design, Make, Test and Analyse are four completely independent stages. But the interfaces between those steps are also vital, and one that often holds back DMTA is the hand-off between Make and Test.
In many pharma companies, the labour-intensive Make stage (at least for early DMTA cycles) was offshored in the early 2000s to keep down staff costs. Although a good decision at the time, the need to ship samples around the world is now increasingly unsustainable, in terms of the delays involved, supply-chain bottlenecks, and the environmental footprint. As a result, the long-term cost-effectiveness of physically placing Make and Test side-by-side is now becoming clear.
Indeed, several leading pharmaceutical organisations are already investing in more integrated discovery environments that bring chemistry, biology and automation teams into closer proximity. While the specific approaches vary, the underlying objective is the same: reducing delays between compound synthesis and biological testing so that decisions can be made faster and with greater confidence.
But while making that change, we should also use the opportunity to rethink compound management workflows across the Make/Test boundary. This is needed because although long-term compound collections are often viewed as the ‘crown jewels’ of pharma, the uncomfortable truth is that most of the compounds in them are likely to be inactive in the majority of assays or screens. This inefficiency is exacerbated by the fact that these compound collections have become centralised to avoid the cost of duplication, further increasing the time and cost of processing compounds through them.
The solution is to synthesise compound libraries on-demand and in-house, starting from commercially available building blocks, and ideally using some of the synthetic paradigms outlined above. The products can then flow directly across to Test, resynthesis and storage is only required for those compounds judged to have long-term interest. Extending this idea further, where the chemistry and biology allows it, you could even test unpurified mixtures, further integrating Make and Test into one streamlined, automation-friendly workflow.
Fresh thinking for the new era of DMTA
The factors driving the transformation of DMTA are already here. AI is increasing the pace of molecular design, while advances in automation and analytics are accelerating decision-making. The question is whether the rest of the discovery workflow can keep pace.
Addressing this challenge will require action on multiple fronts. We need more scalable approaches to purification, more predictive testing systems that can operate at industrial scale, and closer integration between chemistry and biology. None of these changes alone will transform DMTA, but together they have the potential to unlock a step-change in discovery productivity.
Accordingly, the smart players in this space are already coming up with new ideas for how DMTA should work, aided by innovative thinking about everything from synthetic protocols to purification to the use of data in decision-making. In contrast, those that remain committed to legacy processes and structures may find it increasingly difficult to compete. The race to lead the way into the next era of drug discovery is well and truly on.
Ultimately, the organisations that succeed will be those that optimise DMTA not as four separate activities, but as a single learning system.
Ready to increase DMTA iteration velocity?
The next generation of drug discovery will be defined by how effectively organisations connect Design, Make, Test and Analyse into a single learning system.
If you are exploring how AI, automation, data infrastructure or new laboratory technologies could help remove bottlenecks from your DMTA workflow, speak with TTP's drug discovery team about the technologies and operating models that can support higher-velocity discovery.
Get in touch to discuss your DMTA challenges.
About TTP's Drug Discovery Tools Team
TTP helps pharmaceutical and biotechnology companies accelerate drug discovery by removing bottlenecks across the Design-Make-Test-Analyse (DMTA) cycle.
Our Drug Discovery Tools team develops bespoke technologies that enable faster experimentation, higher-quality data and more efficient decision-making. Combining expertise in biology, chemistry, automation, microfluidics, software and instrumentation, we help clients solve complex challenges spanning synthesis, purification, screening, organoid systems, assay development and integrated laboratory workflows.
From targeted workflow improvements to entirely new discovery platforms, we work with clients to increase DMTA iteration velocity and unlock the full potential of AI-enabled drug discovery.
