How the power of Artificial Intelligence saves medicine developers decennia of work

LinkedInTwitterEmailWhatsApp

Even though 90% of all clinical research for new medicine ends in a failure, these studies do gather useful data. They offer a solution while developing other medicine. Perhaps a rejected drug can successfully be used for another treatment, or to boost the effectiveness of other medicine. The problem: there are numerous scientific data. Combing through them is rather labour-intensive. This blog is about how Artificial Intelligence (AI) provides a solution to this problem.

Of course, in medicine development there’s nothing new about researching existing medical literature to make useful connections. The challenge: this takes up a lot of time. It can take years, if not -decades, to search through all relevant documents. This is one of the reasons why traditional medicine development can be so intensive.

Natural Language Understanding (NLU) provides the solution. Medical data analytics company Keystonemab uses IBM’s AI platform to distillate insights from millions of scientific articles. The AI is fuelled by the knowledge and expertise of scientists. They have trained the model by “feeding” it relevant documents. Moreover, they set the conditions for the model to select its data.

AI doesn’t just select the relevant data, but also analyses it

It took Keystonemab a year and a half to train their AI model. Afterwards, the platform is provided with millions of scientific articles, from which the data are automatically extracted. For example, a researcher is looking for data about the effects of a certain molecule on a biomarkers. The AI can select all relevant data. The scientist in turn analyses the insight provided by the model, in order to decide whether they are useful.

The data platform can even play an important role during the analysis phase. Keystonemab namely employs the AI to display promising combinations of drugs. For example, the effects of an existing medicine can be amplified by combing it with another drug. Or perhaps rejected medicine can be used for another treatment.

It works like this: when Keystonemab’s data platform has chosen all relevant data for a certain medicine study, it will automatically display the promising connections. The platform can also estimate the value of these connections, in order to prioritise the possible combinations. Eventually, this leads to an analysis, basis on the scope of the research, including recommendations.

The huge advantage of this approach: it saves a lot of research time, which leads to an sooner start of the clinical trial. If that study has been conducted in an earlier stage (because the combination of drugs is based on already existing treatments), the production phase can be started immediately. This doesn’t only result in a cost advantage, but it also means the new medicine will be available faster.

Would you like to read more about Keystonemab’s progressive approach? Then download the reference case.

Download