Blog: Artificial Intelligence: a new character in the storytelling of drug development and discovery? – PMLiVE
Written by Edward Jones, Junior Medical Writer.
Bedrock is dedicated to creating valuable and powerful stories to engage audiences that drive an improvement in the provision of health. This focus becomes even more important when communicating newly developed and game-changing treatments or therapies that could rapidly improve outcomes. Therefore, while the attention is usually drawn towards the clinical trials or real-world evidence for a new drug, sometimes the developmental story and discovery of such a product are powerful story-telling tools that deserve just as much recognition.
The development and discovery of new drugs often brings to mind the exploration of uncharted rainforests or white-coated scientists conducting research in laboratories. However, in recent decades, artificial intelligence (AI) and machine learning are set to change the rules of the game not least because one of the biggest challenges to pharmaceutical companies is the rising costs of drug development, which have increased 145% increase since 20031, hitting $2.6 billion in 2014.
In essence, AI is the development and use of computer systems that are able to recreate and perform tasks that would normally require human intelligence, such as decision-making, visual perception and speech recognition1. Recent general applications of AI have included the police using it to draw up photo fit drawings of criminals or speech recognition tools that adjust to accents, slang words and even illness by learning from humans as they talk to the device. Machine learning is an integral part of these processes as it allows computers to leverage large sets of data and information without being explicitly programmed and has evolved from the use of pattern recognition in early programmable AIs2. Because of the reliance of machine learning on large data sets, the concept of ‘big data’ has also become a central element in the development of AI. By its very nature, big data can be so immense and complex that it can no longer be analysed by traditional data processing application software and is, therefore, itself a key driver in the development of AI and machine learning.
The rise of big data and machine learning has led to huge strides in the development and practicality of AI in both the corporate and scientific world and, whilst its use is still fairly limited, it is likely to make major changes to the way drugs are developed and utilised. A variety of companies are already using AI to sift through vast amounts of existing quantitative data to find patterns that are then refined over time via deep learning, a subset of machine learning. This allows AI to focus on new information that may have been previously missed by research scientists to improve decision making, discover new drugs and even repurpose existing ones.
For example, one small medical technology company is currently working on its own deep learning technique known as the generative adversarial network (GAN) that uses two competing networks to create outputs that essentially look like real data but are actually ‘imagined’ drug concepts that can be further tested2. Not only is this tool becoming important to drug research and development, it is becoming essential as the sheer number of daily publications becomes too large for science teams to efficiently compile and analyse (around 2.5 million are published each year3). AI could do this job for them, saving a huge amount of time and money whilst making it a more efficient process.
However, despite the recent progress made, there are some limitations. Whilst AI and machine learning are extremely efficient at predicting drug structures and their targets, they are not as proficient at predicting the side effects that may be associated with these new drugs. Even if a drug binds to a single receptor, it can be difficult to predict potential multiple responses (also known as on-target toxicity) which may prove problematic later in the development cycle. Nevertheless, with the advent of personalised medicine and acquisition of more data that allows us to further understand biochemical pathways, AI may eventually be able to solve these problems as well.
Despite early days and some reasonable doubts towards its overall impact, AI is fast becoming a primary focus for many major and minor players in the healthcare and pharmaceutical industries. It draws attention to the increasing complexity and importance of drug development and shows how it can form a central narrative of any engaging medical communications programme. Bedrock understands that in order to engage audiences and understand the needs of our clients, we need to fully grasp the significance of every step in the drugs developmental process. If AI is becoming a central part in the future of drug development and discovery, we need to ensure we fully communicate its implications and value. It is only then that we can really understand the thematic messaging behind its story and tailor medical communications programmes to deliver the greatest impact possible.
To find out how Bedrock Imagines, Creates and Delivers visit us http://www.bedrock-health.com/
- V. Eremenko, ‘What is artificial intelligence?’, 2015, BBC, http://www.bbc.co.uk/news/av/technology-34224406/what-is-artificial-intelligence [accessed 08/01/2018]
- V. Maini, ‘Machine Learning for Humans’, 2017, Medium, https://medium.com/machine-learning-for-humans/why-machine-learning-matters-6164faf1df12 [Accessed 08/01/2018]
- R. Mullin, ‘Cost to Develop New Pharmaceutical Drug Now Exceeds $2.5B’, 2014, Scientific American, https://www.scientificamerican.com/article/cost-to-develop-new-pharmaceutical-drug-now-exceeds-2-5b/ [accessed 03/01/2018]
- S. Hill, ‘How AI Could Help Reduce the Cost of Drug Discovery’, 2017, LEAF Science, https://www.leafscience.org/ai-and-research/ [accessed 04/01/2018]
- S. Boon, ‘21st Century Science Overload’, 2016, Canadian Science Publishing, http://www.cdnsciencepub.com/blog/21st-century-science-overload.aspx [accessed 03/01/2018]