Even though new medicines, therapies, instruments, and methods are founded and introduced each year, contemporary medicine still faces many problems and lost opportunities. Medical and scientific communities hope for artificial intelligence, big data technologies, and machine learning to resolve unsolved problems.
New Medicines
Designing and introducing a new medicine requires time and high expenses. While some compounds are quickly approved, others are rejected. Although some are made through the selection and become drug candidates, they are eventually of low efficiency.
Algorithms of deep learning can simplify (to some extent) the search for new drugs and improve their efficiency. This approach makes specially trained neural networks find the compound that has all chances to become the new medicine base. Then, these networks can analyze an extensive database of those candidates that can make the perfect match.
Insilico Medicine, founded in 2014, is thought to be the pioneer of AI usage for drug invention and development. The company used powerful computers for designing small molecules having specific features that could be the base of the new medicine.
The research on GENTRL neural network is a famous example of the advantages suggested by AI assistance. The network needed 21 days to model potential inhibitors of DDR1, the protein implicated in fibrosis and some other diseases. Afterward, the project spent some time synthesizing the molecules.
One of the main points expected to be solved by AI assistance is discovering new antibiotics. However, the World Health Organization (WHO) is alarmed and points to the fact that no new antibiotics have been discovered in recent years, and only modified versions of medicines discovered in the 1980s are widely used. Worse, an increasing number of infections are caused by superbugs that are strong and resistant to most antibiotics, if not all of them.
The Organization encourages scientists to actively participate in developing new antibiotics before the existing ones cease being efficient.
Scientists are trying to invent new antibiotics by using deep learning methods. Research stated that Halicin, a potential broad-spectrum antibiotic, had been found in the Drug Repurposing Hub database in 2020. However, Halicin alone cannot solve the antibiotic resistance problem. AI is expected to be used for determining other potential compounds.
One more medicine that is awaited desperately is the medicine that can slow down or stop the progression of Alzheimer’s disease. Alzheimer’s disease is the most widespread type of dementia that has become the trouble of our time because life expectancy continues to increase. According to the estimations of the WHO experts, in 2021, 55 million people will have dementia. By 2050 the number of patients is expected to reach 139 million. Currently, scientists do not know what causes Alzheimer’s disease, and doctors have no medicines to cure this disease.
There is only one medicine of this type that the FDA has approved over the recent two decades, called aducanumab, which Biogen developed. This medicine is based on monoclonal antibodies and aims to destroy beta-amyloid plaques that clump in the brain.
Despite the impressive results obtained at the initial stage, the clinical trials ceased in 2019. The experts concluded this medicine was inefficient. As a result, the manufacturer had to re-initiate the procedure for approval. It stated that higher doses would show a better effect. The regulator actively supported the decision.
Also, the European Medicines Agency denied authorizing the medicine on the market because of deep concerns over its low efficiency and possible side effects (confusion, brain swelling, and tremor).
Clinical trials of medicines for Alzheimer’s are still in progress. Although many substances that showed some efficiency happened to be frustrated during the final stages. One of the examples is the case of dimension. Medicines researchers put their hopes on this medicine. Atuzaginstat is another example of the expectations that failed.
One more principal barrier is that the invention of medicines for Alzheimer’s is one of the most highly priced in the industry. On average, this method takes 13 years and costs a minimum of $ 5.6 billion.
Researchers think that artificial intelligence is going to become the main differentiator. Therefore, they are actively implementing it for making a diagnosis and finding medicines for Alzheimer’s disease. New ideas cast doubt on the amyloid hypothesis of its pathogenesis, and it is necessary to ensure diagnostic accuracy and search for efficient medicines. Moreover, artificial intelligence technologies can speed up this search and make it less cost-consuming. Therefore, we undoubtedly need to get efficient medicines here and now.
Exscientia is a British AI medicine discovery company. It believes the medication based on the DSP-0038 substance can solve the problem with the disease. The business tested its hypothesis on the compound’s efficiency against Alzheimer’s disease on the platform of artificial intelligence called Centaur Chemist. The clinical trial of DSP-0038 started in the US in 2021; the test was a great success, and artificial intelligence proved to have solved the problem.
One more tendency that has been popular in recent years involves using existing medicines for curing other nosologies rather than they were developed for; in other words, they are getting the status of “multipurpose.”
This re-purposing simplifies and makes the invention of new therapeutic approaches quicker and more cost-effective.
Artificial intelligence eases the process of resolving those tasks because the latter, until recently, relied mainly on intuitive suggestions rather than on a systemic approach. Healx launched one of the most important re-purposing-based projects. This company intends to find re-purposed medicines that could cure 7,000 rare diseases, where only 5% of them have an approved treatment.
Anti-Aging Medicines
The life expectancy goes on to grow. This growth makes the “aging as a disease” concept more and more popular. Many people do not want to accept age-related health deterioration and wish to stay young.
Those who follow this idea (biohackers) use the most unusual strategies, e.g., adjusting their diet, strictly regulating the sleep-wake cycle, and taking medicines that could help them live longer.
The trend of using off-label medicines that have already been introduced to clinical practice is also gaining popularity. One of these medicines is rapamycin. It was initially administered to patients who recovered from organ transplantation. Rapamycin was tested on flies and mice and displayed a rejuvenating effect — one mouse lived for three years and eight months, twice the average lifespan of these rodents. Although rapamycin has not been tested on people yet, this is what is expected soon.
One more example is the medicine called metformin, a popular anti-diabetes drug. According to the studies, it improves the body’s response to insulin, positively influences blood vessels, and provides a rejuvenating effect. In the US, metformin can be accessed even by patients who do not have diabetes because AgelessRX offers prescriptions for this and other anti-aging medicines.
Besides, the demand for a universal pill to live longer is still extremely high. Dozens of pharmaceutical companies are attempting to develop efficient medicine. A lot of them use technologies of the artificial intellect. In 2021, scientists from Surrey University stated their ML models that analyze the DrugAge database helped find at least three components that were able to delay the onset of aging. These are famous antioxidant flavonoids, omega-3 fatty acids, and organic oxygen compounds. Does this mean that these three components will become ingredients of a “super anti-aging pill”? We will see what time will tell.
Health Monitoring
The tendency for health self-monitoring by using wearable devices arose several years ago. Such devices include Holter monitors, BP meters, smartwatches, fitness trackers, and even tattoos that can monitor, for instance, blood glucose.
Some medical-related technology can send their data directly to the user’s smartphone or the doctor’s computer. What is important most of all is not the data itself but the potential results of analyzing it. Artificial intelligence can process all the collected information, e.g., Apple’s Health Kit. The data sent by wearable devices is analyzed and recorded in the patient’s EHR in the most convenient format.
A perfect artificial intelligence (the one that can become a peer or at least a highly qualified assistant of human doctors) still needs some improvements. Then, the AI can put it in charge of concluding the data sent by wearable devices, estimating the risk of some diseases, searching for an efficient dosage regimen, ordering tests, and, in case it is necessary, requesting to call for emergency assistance. Finally, it results in the fact that the solutions provided by artificial intelligence will be based on a large array of data and tailored for individual patients.
Artificial intelligence is critical in triage, i.e., the initial sorting of patients who request medical assistance. Triaging divides patients into groups depending on their features and the assistance required by them.
Artificial intelligence can already reduce the workload on medical staff and perform triage in all areas of medicine. For instance, it can analyze mammography screening results or estimate the physical and mental state of COVID-19 patients.
Importantly, it is essential to remember that people are not ready to trust artificial intelligence if it makes a diagnosis because a computer cannot take into account and consider some relevant factors for diagnostics and therapy. Thus, it will probably encounter difficulties in learning the principles of medical ethics. Also, this particular roadblock is why computer assistants (regardless of how much they are demanded) can never substitute qualified human doctors: the latter will always have the final word.
Comprehensive Diagnostics
Despite the success in several medical areas, diagnosing many diseases is still a crucial and challenging issue. This challenge is where artificial intelligence can come to the rescue.
The clinical practice already uses the technologies that can make a better diagnosis of cancer. For example, neural networks are widely used to diagnose melanomas. A comparative study in 2018 showed that the diagnostic accuracy of artificial intelligence was 95%, while a highly qualified diagnostician identified melanoma only in 90% of all cases. The same approaches are used to diagnose ophthalmological disorders, various types of cancer, and many other diseases.
However, a universal check-up can become a good solution for many patients. They may avoid going to different specialists but take a blood test that comprehensively analyzes their health status and potential risks. Moreover, this technology can entirely change the diagnosis as it is known.
Scientists are already making progress in this area. A 2020 paper describes an all-encompassing cancer test; the authors state that only one blood sample can help diagnose 50 types of cancer. The artificial intelligence trained on the Circulating Cell-free Genome Atlas database makes the diagnosis. Its accuracy may vary depending on the progression of the disease: the later the stage was, the more precise the diagnosis (93% for Stage IV and only 18% for Stage I) was. There is no doubt that a similar analysis may soon become a clinical routine, as well as the CBC test, as long as they get more precise.
Moreover, neural networks can be much helpful in diagnosing mental disorders, where the main symptoms are defined by behavioral changes rather than by laboratory indicators.
One of the pressing issues that today’s medicine face is disorders of the autism spectrum. But, according to specialists, it is rather critical to identify the disease as early as possible: the sooner therapy starts, the more efficient it will be.
Scientists from Geneva suggested a diagnostic option based on artificial intelligence. Their system analyzes 10-minute videos that show a child who plays with a parent or a legal career. The diagnostics method focuses on children below 5: this is the age when it is complicated to identify autism.
Autism can have various symptoms, and the system developed by Swiss scientists concentrates on definite motion patterns related to the disease. The neural network, trained on video clips showing children with confirmed ASD, displayed 81% accuracy. The authors of the invention state this is a considerably high level. Artificial intelligence can become an assistant or a supporting instrument for doctors who work with children and their parents. It could become easier to diagnose, and Artificial Intelligence could prescribe the proper therapy on time.
Generally speaking, it is possible to expect that in the future, medicines against diseases that current science cannot currently treat will be found, diagnostic methods will be more advanced, and treatment will be even more individual. Furthermore, since a vast amount of data gets collected and accumulated, and special-trained neural networks are increasingly used in a clinical routine, it will be much easier to diagnose diseases and make prescriptions for treating certain patients. It may sound too good to be accurate at the moment, but the future will make it accurate.
About the Author
Rustam Gilfanov is a LongeVC investment fund venture partner and a business angel.
