Is augmented intelligence the key to achieving true digital transformation in pharma? Peter Crane outlines the need for a scientist-centric approach
Technology has dramatically altered the global landscape, with incumbents in sectors such as retail, travel and hospitality all challenged by the emergence of technology-enabled businesses. The biopharmaceutical industry, however, has yet to be significantly challenged, despite tech giants making overtures towards the broader healthcare sector.
Scientific and regulatory complexity, as well as the long- term nature of drug R&D programmes, impedes the ‘move fast and break things’ approach adopted by technology companies in other industries. In spite of this, there remains a growing opportunity for digital innovators to effect positive change within a biopharmaceutical R&D environment. The difference is that, rather than ‘disrupting’ pharma, technology products should amplify the inherent strengths of biopharma companies to afford real improvements in R&D productivity.
The greatest strength of the biopharmaceutical industry is in its people – the highly qualified scientists who design, develop and test new therapeutics. For a new technology to deliver a tangible benefit, it should allow this pool of talent to spend more of its time focused on the innovative aspects to its roles. In particular, within the lab, it should mitigate the tedious and error-prone manipulation of biological samples, as well as allowing researchers to work comfortably with complex experiments. For the average bench scientists, this could simply mean allowing them to spend less time in the lab preparing plates for routine assays or saving them hours when trying to collate and analyse large quantities of data from an imaging run.
There is also the move towards increasingly complex modalities such as cell and gene therapies, as well as a growing adoption of high content screening. These complex living modalities are challenging the status quo of manufacturing, as their high degree of variability and increased analytical burden are contributing to a high COG. This activity is driving interest in data-driven adaptive cell processing. Likewise, multidimensional high content screens are causing challenges to researchers tasked with analysing and interpreting the vast quantities of data produced. In both these instances, the movement away from reductionism is contributing to a data deluge, which will overwhelm an R&D infrastructure built around a reductionist approach for biology.
Within the biopharma landscape, the introduction of a new toolset is required which will empower biopharma scientists to work in ways compatible with complex, high dimensional and data rich systems. This toolkit will enable companies working in advanced therapies to lower the costs and expand their labels, whilst in traditional biopharma incentivising scientific quality over quantity has already contributed to impressive improvements in R&D productivity (eg the 5R framework of AstraZeneca).
The toolkit will initially be assisted intelligence quickly followed by augmented intelligence, rather than true artificial (autonomous) intelligence. In a pharma environment, this is at present often just an external contract research style engagement. Within the lab environment, for example, a system that is introduced to enable researchers to intelligently automate the collection, structuring and collation of multiple data streams may be classified as assistive. This system enables researchers to do existing tasks better or faster. The second stage is towards augmented intelligence and the partially closed loop of design, build, test and learn (also known as Computer Aided Biology). In this system, data is collated and structured from multiple sources and reviewed by the researcher, but the difference is that now the system is able to enact actual process or experimental changes it has inferred from human interaction (active learning).
Implemented properly, this will enable previously unobtainable levels of process or experimental control. Here we see industry interest in areas such as adaptive control in bioprocessing and digitally powered screening platforms in early drug and therapeutics discovery. At the most extreme, we have a fully autonomous (or a fully closed loop) system in which the software learns and is able to enact the next process step autonomously in a continuous cycle of improvement. Arguably this vision is some way off in a heterogenous biopharmaceutical R&D lab environment but may be closer to realisation in a lower process complexity system such as a DNA foundry.
To see positive changes within biopharmaceutical R&D, we need to solve the productivity challenges that are impeding progress today but in a way that naturally tends towards more augmented systems and empowered teams. The increase in data complexity and quantity represents an immediate challenge to organisations. To address the data deluge in R&D, we must fix the challenge at source – in the lab.
Given that bench scientists spend inordinate amounts of time preparing complex assays, whilst data scientists can spend up to 80% of their time cleaning data for further analysis; technologies that can augment researchers and encourage digital ways of working within the lab will deliver wider benefits. Digital solutions on their own will not deliver the productivity gains the industry seeks, only by introducing tools that empower R&D teams to work more efficiently in the laboratory setting will the true power of data science in biopharma be realised.
Dr Peter Crane is a consultant at Synthace (www.synthace.com)