Machine learning – the technology which allows machines to learn from data and self-improve – has significant potential for advancing a range of industries, as we’ve previously discussed in relation to manufacturing and the professions.
On 8 September, the Royal Society held a workshop on the use of machine learning in the UK pharmaceutical sector. We explored what machine learning could do, and barriers to its use, with global pharmaceutical companies, sector associations, regulators, start-ups and SMEs. This blog post gives a few highlights from the discussion.
What can machine learning do for the pharmaceutical sector?
Its ability to spot patterns in large volumes of data gives machine learning a range of applications in the pharmaceutical sector. For example:
- In bio-manufacturing, machine learning can be used to optimise processes, using data produced from experimental or manufacturing processes to reduce the time taken for a procedure, reduce cost, or improve reproducibility.
- In early-stage drug discovery machine learning can sift through vast amounts of data to detect patterns that elucidate the complex biological systems at work. For example, machine learning could be used in the initial screening of drug compounds to predict success in compound activity and interaction.
- In personalised medicine, machine learning could use patient data to predict responses to particular treatment pathways, enabling tailored treatments which are more effective.
What are the barriers to reaching this potential?
Data is the fuel for machine learning yet collecting and accessing useful data can be difficult in the pharmaceutical sector: in many cases, the most useful data is personal medical data, which is unlikely to be shared. New apps and wearables such as FitBits could also provide medical data that could be used in new ways. If machine learning is to be put to use in pharmaceutical and healthcare applications – and we are already seeing machine learning companies move into healthcare – these data governance issues will need to be addressed. Our data governance work with the British Academy is therefore exploring the governance of data in relation to its uses.
Regulation can be a barrier in the pharmaceutical industry as it takes time, effort and money to verify a product and bring it to market. For machine learning, stringent regulations in drug development can be particularly challenging due to the ‘black box’ nature of some algorithms which cannot provide causal explanations for their results. This lack of transparency could affect uptake, even if the results are highly accurate.
As we found in our manufacturing roundtable, pharmaceutical companies also have difficulty recruiting individuals with the right data science skills.
The benefits of machine learning in the pharmaceutical sector are potentially significant, from day-to-day operational efficiencies to significant improvements in human health and welfare arising from improving drug discovery or personalising medicine.
Achieving these benefits requires addressing some difficult issues around building a skills pipeline, regulations for the use of data or for product innovation, altering business practices, and improving awareness. However, GlaxoSmithKline’s recent regulatory approval for stem cell gene therapy shows that barriers can be overcome and pharmaceutical companies can make transformative advancements.
In machine learning specifically, advances are also being made. For example, machine learning is being used in a personalised treatment for depression, analysing a range of data to assess whether a particular treatment pathway is likely to be successful for a patient. Further ‘proof of concept’ case studies – showing where machine learning has had a major business impact in the pharmaceutical sector – could raise awareness and encourage further dialogue.
The Royal Society is currently carrying out a project on machine learning, which is considering the potential of this technology, and the challenges that come with it. You can read more about machine learning in this interview with Marcus du Sautoy.