Playing by the new rules
Intelligent business design
Tatiana Kublitskaya, Pharmaceutical Bulletin
PharmaTimes portal discussed the future of artificial intelligence technologies in relation to research and development with AI expert Jackie Hunter.
Ms. Hunter, CEO of BenevolentAI, believes that, thanks to the use of artificial intelligence (AI) technologies, the list of TOP20 pharmaceutical companies will undergo significant changes over the next five years.
Speaking at a recent briefing, Ms. Hunter noted that the current list of TOP20 players in the global pharmaceutical market is significantly different from the one that could be seen 10 years ago. And if you take a similar list from twenty years ago, there were no players like Teva and Gilead in it at all. The biotechnological revolution and a new understanding of diseases have become the drivers of change. Therefore, according to Ms. Hunter, nothing can prevent artificial intelligence from becoming a driver of such changes. One can even expect a company like Google to appear in the future TOP20 list, which has already made significant investments in the development of biotechnologies and diagnostic tools and has very good positions to get into the "upper echelons" of pharmaceutical market players.
AI is a rapidly developing trend in pharma. Several representatives of Big Pharma, including Merck & Co (outside the USA and Canada – MSD), Johnson & Johnson, Pfizer and Sanofi, are already conducting research in this area. For example, GSK recently signed an agreement worth 33 million pounds with Exscientia from the Scottish city of Dundee, specializing in the field of AI. The purpose of the agreement is partnership in the field of discovery of new medicines.
However, despite this, Ms. Hunter is not aware of pharmaceutical companies that have a comprehensive AI strategy, although, in her opinion, this would help debug the "unstable" model of finding and discovering new molecules. Currently, the proportion of failed phase II and III clinical trials is 50%.
The BenevolentAI platform works mainly based on data analysis and the search for links between them. For example, in the field of motor neuron disease / amyotrophic lateral sclerosis (ALS), the AI solution developed by the company was able to analyze billions of sentences and paragraphs of scientific articles and abstracts to establish links between data that were classified as "known facts" with the help of this solution. These facts were examined by specialists, which allowed creating new connections for generating a large number of possible hypotheses based on the criteria set by scientists. In the case of ALS, there were about 200 such hypotheses.
Then the researchers could evaluate the validity of the hypotheses and make a priority list of those that made sense to develop further. In the end, five hypotheses were selected for testing in the laboratory. In general, the test results were positive: of the selected candidate molecules, one demonstrated an effect on ALS models in vivo; two showed excellent efficacy in quantitative analysis; two demonstrated a smaller, but significant effect; and one was ineffective.
That is why Ms. Hunter does not believe that the introduction of AI will lead to massive layoffs in the field of clinical research, and argues that the realities of this technology are actually closer to "augmented" than to artificial intelligence.
She notes that in any case, researchers will have to use their own intelligence and experience to test the generated hypotheses, but they will receive a much more extensive evidence base with fewer deviations.
It is also an important component in maintaining the patentability of the drug. To ensure patentability, the scientist interprets the data using situationally determined methods, since with full automation of all the processes of searching for molecules, it will be unclear what the invention actually consists of.
To find links between data, BenevolentAI had to create a database that includes tens of millions of publications that can be read using AI programs. The company uses a huge number of data sources – both structured and unstructured. The problem may be data disorder, and the company will need 2 years to create thematic dictionaries that will overcome this problem. For example, the abbreviation AD may refer to atopic dermatitis or Alzheimer's disease, so AI must understand the context in which this abbreviation occurs.
Ms. Hunter also says that AI can help overcome the problem of disparate data. For example, an array of oncology data may contain information about antimicrobial resistance, or information about Alzheimer's disease may also include information about atopic dermatitis, but in conditions where AI is not used to get a broader picture, the data is usually organized in the form of disparate arrays.
At the initial stage, companies may face the following problem: AI technicians who have never worked with pharma will have to interact with clinical researchers who have never dealt with AI. This problem also existed for BenevolentAI, but the company found the following solution: people were encouraged to work together, setting common goals for them; at the same time, they proceeded from the understanding that the culture of the organization should be changed to ensure more effective communications.
New methods of searching for new molecules using AI may also force regulators to change the way drugs are evaluated, although Ms. Hunter believes that regulators are far behind in understanding such technologies. She is also sure that Big Pharma companies should adapt to these technologies faster so that they are not "overtaken" by startups that are more receptive to innovation. She says that large companies can buy startups using AI; however, if they themselves do not have internal systems to maximize the achievements of startups, they will simply "kill the chicken that lays golden eggs." We need an environment in which AI startups can develop, and not overcome the resistance of internal structures.
In general, Ms. Hunter believes that it will take no more than 10 years before the use of AI in research and development will become the norm, not the exception.
Portal "Eternal youth" http://vechnayamolodost.ru
22.09.2017