AI in Drug Discovery Trends
Linkedin | GitHub
Having worked the past 3 years as a data analyst for a life science startup accelerator, I was interested in using Python data visualization tools to discover trends in healthcare startups. I was able to find data in two BenSci blogs that focused specifically on startups working with AI in drug discovery.
Drug discovery aims at identifying drugs that prevent or treat a disease. Currently, the estimated cost of bringing a new drug to market is $2.6 bn, and only a small percentage of the drugs that enter clinical trials are approved (10% in 2015-2016). This leads no one to doubt that there is a serious unmet need for innovation in drug discovery.
Researchers are looking at AI as a way to accelerate and reduce the cost of discovering new drugs. AI systems allow researchers to extract insights from huge datasets. Use cases include predicting the properties of a potential compound, generating ideas for an entirely novel compound, and alleviating the need for repetitive tasks.
The Data
I used Scrapy to scrape data from BenchSci, a Toronto based company that uses AI to empower scientists to run more successful experiments to accelerate drug discovery.
Simon Smith, CSO at BenchSci, wrote the two blog posts I pulled data from. The first dataset focused on 116 unique Drugs in the AI in the Drug Discovery pipeline featuring 262 unique variations of the drugs. The second dataset listed 230 startups using AI in Drug Discovery.
My goal was to use descriptive analytics to try to understand and describe AI in Drug Discovery startup trends in the US and globally.
Results
The top five countries with the most startups in the dataset were the United States, the United Kingdom, Canada, France, and China. The yearly number of startups using ai for drug discovery by country is shown below. The graph shows the number peaked in the US in 2017, the UK in 2012, Canada in 2015, France in 2013-2016, and China in 2017.
The box plot below also shows that the US has the largest range of founding dates based on the datasets, having the earliest and the most recent startup. When sorted by year founded you can see that the median year for each of the top 5 countries was between 2014-2017.
Most drugs listed were in the preclinical or discovery phases. The top three therapeutic areas of the drugs using AI in drug discovery were oncology, neurology, and infectious disease with oncology being the most common therapeutic area. This tracks with my past experience with healthcare startups. There has been more funding, whether it be angel investors or foundations, investing in cancer research. Of the countries that have startups in all three areas, oncology does lead, with the exception of France, where infectious disease is more common.
Future Work
For future work, I would like to have a more in-depth look at the impact of these startups. I could do so by first looking at their funding, both private and public sources. The number of jobs created and team size would be interesting to see from an economics perspective to see if these startups are creating new jobs. Knowing if the startups are working out of a coworking space or incubator or need space for manufacturing could also provide further insight on their potential real estate impact.
About Author
