A recent analysis from Forbes shows that the pharmaceutical industry has a declining business model.
The study goes into details and explains the situation: The low-hanging fruit of the pharmaceutical sector has already been picked, simple vaccines and generic drugs were invented over decades ago, which means that discovering new drugs requires consistent investment in R&D.
However, the money for R&D comes from available revenue. As people become more and more healthy, they spend less money on drugs. Therefore, there is less money for developing new drugs. In this situation, a change of paradigm, as well as a change of the methods, is necessary. We could say that we are now facing the most costly and hard to solve problems in the pharmaceutical industry.
AI development seems to be a way out through efficiency and automation. Its predictive power offers increased precision and the ability to save time and money on research.
However, the fact that AI development is still in its infancy and there are very few experts in the field represent important obstacles.
There are several ways to use AI for the pharmaceutical sector which we will discuss in the following sections.
New drug discovery and clinical trial research
As pointed out in the introduction, creating new drugs requires substantial amounts of money for research and development since this is a trial-and-error process with high failure rates.
AI and specifically machine learning can perform an initial screening of potential drugs and predict successfully if these are possible cures to existing diseases. The great advantage of this approach is that it rules out quickly unsuccessful compositions and that it allows doctors to gather information on substances that work much quicker. New drug development is based mostly on unsupervised learning and has been used so far by Microsoft, focusing on cancer and the MIT studying type-2 diabetes, as well as The Royal Society for drug optimization.AI, also is also useful for clinical trial research, not only creating drugs. For example, the pattern-identifying capacities of the AI system can be put to good use in finding the right candidates for a new drug. This can reduce the number of hours research assistants spend on work, replacing their efforts with a simple scan of an existing database.
AI, also is also useful for clinical trial research, not only creating drugs. For example, the pattern-identifying capacities of the AI system can be put to good use in finding the right candidates for a new drug. This can reduce the number of hours research assistants spend on work, replacing their efforts with a simple scan of an existing database.
The ML system can automatically compute the necessary size of the testing sample and recruit the best candidates by taking into consideration not only the disease the study focuses on but also secondary conditions which could have an impact on the overall evaluation of the drug.
Until now, the medical process looked at particular cases and tried to find the general rule. When this was identified, the treatment was aimed towards the general condition, resulting in important side effects related to the particularities of the patients.
AI offers the opportunity to reverse this process and find drugs which work for a particular individual, looking at that person's medical record and considering the smallest details.
When this becomes the norm, it will mean less allergic reactions and far fewer side effects.
Since AI is very good at identifying patterns, it can be used for diagnosing diseases and combined with the predictive precision medicine previously mentioned.
So far, there have been significant advancements regarding cancer treatment, by using IBM's Watson AI engine. The project aims to sequence the genome and identify those genes which can cause cancer.
Google is not far behind with its project DeepMind Health which focuses on treating the degeneration of aging eyes.
It is expected that AI systems will be able to diagnose patients much quicker than their human counterparts, as well as uncovering more subtle relationships between symptoms and diseases, thus accurately diagnosing complex conditions.
Better electronic health records
Electronic health records are a collection of unstructured data ranging from medical observations, to test results, X-rays, and other items. Some of this information is handwritten, which makes it even more difficult to find specific data such as the response to treatment, quickly.
In these conditions, it is difficult to classify the contents of the record and to use it for patient selection or other medical services.
Now is the correct time to introduce intelligent, electronic health records which a machine can automatically scan for things like the given diagnosis, the drugs used, and the patient's response to such treatment. ML can make sense of handwritten data, and optical recognition can help digitize existing records such as medical tests and lab analysis.
Guarding the health of the population
ML systems can do more than individual analysis of health records. They can be used to scan entire populations and predict the risk of disease outbreaks.
This is especially important in the developing world where they may not have the necessary tools to fight epidemics due to a lack of resources and infrastructure.
The prediction of such an outbreak is usually made by looking at real-time data from the web, satellites, social media and other sources.
HealthMap is a project which aims to automatically classify and create visualizations of every state's health situation.
Add ML becomes more of a reality and less of a trendy buzzword, the pharmaceutical industry will adopt this tool as a way to survive.
AI offers a way to optimize processes with minimal costs and the ability to replicate results.
When it comes to creating new drugs, AI can save thousands of dollars on research and also time, by ruling out through pattern analysis those combinations which would not produce a winning result. Used just for initial screening, AI is still a better option than current practices.
The applications do not stop there but expand into personalized medicine which represents a new paradigm of treating diseases, focusing on tailor-made treatments.
Although this method has enormous potential for the pharmaceutical sector, it is essential to think about the obstacles. In this case, personal data privacy and other regulations could be a potential barrier against development. All companies considering these tools need to find ways to comply with the security requirements and also have enough material to train the algorithms. Finding a skilled workforce in AI is also a challenge, as the technology is too recent and there are few educational programs focusing on this area, most experts are self-trained in AI coming from an engineering, statistics or mathematical background.