Pharma Lab: What are the key trends you’re seeing across the drug discovery sector at present?
Jon Wingfield: Everyone is talking about the ability to link automation islands to deliver end to end automation.
Historically, automation within life sciences has really referred to the assembly of assay reagents – liquid handling systems that can build reagents into test wells of a microtiter plate and gather data on the ability of test samples to affect the biology within a test.
With the realisation that efficiency needs to improve across pharma and biotech, the industry is adopting modern technology such as AI, machine learning, and enhanced data analytical tools to connect isolated islands of automation to enable the automation of workflows, not just assay assembly.
The ability to follow samples through cascades of testing and to be able to quickly analyse complex data sets has improved the ‘screener’s’ ability to make accurate rapid decisions within projects. The aim is to reduce the critical timeline between identification of active ‘hits’ to the development of clinical candidates.
The real value of automation is the ability to generate reproducible high-quality data over this extended testing period and not in high throughput screening (HTS).
While automation was traditionally employed to support HTS because of the high volume of testing the value of that data was actually low, since all you find in HTS campaigns are starting points for drug discovery not drugs.
Higher value data is actually generated later in the discovery process during the design, make test analyse (DMTA) cycle when hits are turned into leads then drug candidates. This DMTA cycle time can be years so reproducibility is key – being able to run the same assay week in week out and generate the same data, regardless of the experimenter or the batches of reagents used.
There is strong interest in automating chemistry which has lagged behind biological automation. Unlike biology where there has been the SBS standard microtitre plate and testing is done using simple water-based reagents, chemistry automation is more challenging as there is little standardisation and the solvents are harder to work with.
It is clear that if you are going to make a difference and reduce the DMTA cycle time, then you really need to have the make and test components of the cycle carried out as close to each other as possible.
Pharma Lab: What are the main challenges you believe the sector will face in the next three to five years?
Jon Wingfield: Chemistry automation remains challenging and the outsourcing of chemistry to the Far East for economic reasons is now slowing down the whole Design Make Test Analyse (DMTA) cycle.
Around 10 years ago many western pharma companies started to outsource chemistry to areas of the world where the resourcing costs were lower. At the time they were less concerned with speed and cost saving was the focus. Now with various geopolitical changes and the realisation that time to market is a higher priority it has become challenging to deliver speed. In addition, the environmental impact of shipping samples around the world is not trivial. Integrating automation and driving digital transformation is not without its challenges.
At the basic level it can be very hard to integrate instruments when the suppliers of the equipment have no real interest in supporting integration with other suppliers’ hardware. There is huge potential for improving automation in cell screening, we are starting to see companies emerging who have solid products that can meet the needs of routine cell culture.
There is more that can be done to support automation of more complex cell model systems:
• Better cell models of disease will generate better data and inform better decisions
• Enabling improved cell models earlier in the efficacy testing cascades would be challenging but highly valuable.
Pharma Lab: How do you see automation evolving to address these challenges?
Jon Wingfield: We are starting to see a transition towards automated chemistry. There seems to be two approaches being taken. The first is to build automation that is essentially capable of replicating what a human synthetic chemist can do, two-armed automation systems that can work inside a standard chemistry hood. The other approach is to automate chemistry synthesis at small scale, then make what you need to test, then only scale up actives. Challenges come when purification is required, particularly when there is very little material generated in the synthesis.
Direct to biology testing is gaining some forward momentum. With small scale synthesis using very reactive reagents, progress of the reaction can be assured.
Challenges come when purification is required, particularly when there is very little material generated in the synthesis.
Since the adoption of cryopreservation, HTS no longer needs cell culture automation that can routinely culture hundreds of plates of cells per day – rather the application in DMTA is likely to be a requirement for routine cell culture of 15-20 cell lines but only five-10 plates of each cell line for testing. This presents challenges for automation and the scheduling software of the system.
Pharma Lab: Where do you see AI having the biggest impact in drug discovery, and what are the main limitations here?
Jon Wingfield: AI is having a big impact in cell screening where image analysis and complex data sets are becoming more common. The ability to be able to predict the outcome of a cellular assay within 24 hours as opposed to five-six days would be a significant advantage.
Using the latest techniques, it is possible to apply AI to monitor cells over time and exposure to samples. By knowing the outcome of the test samples and capturing the subtle early changes within the cells, it may be possible to accurately predict the outcome of a five-day assay in 24 hours.
Now we are seeing companies using AI tools to take written protocols and convert them into code to drive automation.
It may soon be possible for someone with no automation experience to ‘vibe code’ a protocol on a complex automation platform, just by telling the machine what experiment needs to be done.
