table of contents
- What is AI?
- Background to the attention being paid to the use of AI in drug discovery
- “Deep learning” is the key to using AI in drug discovery
- Challenges in using AI for drug discovery
- Movements by companies, universities, etc. regarding the use of AI in drug discovery
What is AI?
First, let’s briefly review “AI”. AI is an acronym for “Artificial Intelligence, ” which is a combination of “ Artificial ,” meaning artificial, and “Intelligence,” meaning intelligence, and is translated as “Artificial Intelligence” in Japanese. The Japanese Society for Artificial Intelligence, which was established for the purpose of researching artificial intelligence and disseminating knowledge, defines this artificial intelligence as “a technology that aims to accurately perform advanced inferences on large amounts of knowledge data.” In other words, AI is “computer technology that analyzes large amounts of data to make predictions and find ways to solve problems.” The use of AI has begun in all fields, and it has already become an indispensable part of our lives.
Background to the attention being paid to the use of AI in drug discovery
AI is also attracting attention in the field of drug discovery, and there are great expectations for it. The reason for this is that the introduction of AI can dramatically accelerate the speed of new drug development, significantly shorten the development period, and reduce the costs required for development.
To develop a new drug, we first discover the proteins in the body that are the drug’s targets, and then search for substances that are drug candidates. Once the active ingredient compound has been optimized and passed animal testing and clinical trials for humans, the drug can be submitted for application and approval.
However, there are tens of thousands of substances that can be drug candidates in the libraries of pharmaceutical companies, and there are over 100,000 types of proteins in living organisms, so it is extremely difficult to find the optimal combination among them. It’s work. It is said that drugs with relatively simple structures using low-molecular compounds have already been developed, and the probability of success in developing a new drug today is less than 1/25,000, and the development period is over 10 years. It is said that the cost will be over 100 billion yen.
AI drug discovery using AI is expected to be effective in the development of new drugs, which is extremely difficult. By relying on AI to perform calculations such as predicting compound binding by matching library compounds and virtual compounds that do not yet exist with in-vivo proteins and data from the 3 billion human genome one by one, we are able to solve problems that were previously unthinkable. It has become possible to search for and optimize compounds in a short period of time and at a low cost that would otherwise be impossible.
“Deep learning” is the key to using AI in drug discovery
The key technologies for drug discovery using AI are “machine learning” to realize AI, and “deep learning” which is one of its methods.
AI learns a large amount of examples in advance and matches the best one from the learned patterns. ” is called. Among such machine learning, “deep learning” is created by imitating the functioning of human brain neurons.
In general machine learning, humans input “hints” for learning. For example, if you want to train a machine to identify images of “dogs,” you would feed it a large amount of image data with the tag “dog” and let it learn.
In contrast, with deep learning, the machine is able to find “hints” on its own about what to learn without any human help.
Deep learning, which allows computers to identify problems on their own, has made it possible to discover minute differences and characteristics that humans cannot notice, and to open up fields that could not be realized by humans. This is the reason why “AI drug discovery” that utilizes deep learning technology is attracting attention.
Challenges in using AI for drug discovery
AI has great expectations as a tool for new drug development, but on the other hand, issues with AI drug discovery have also been pointed out.
In order for AI to demonstrate its power, a large amount of data is required for learning. However, currently there is not an environment in place to fully utilize medical data for AI drug discovery.
Medical data held by individual medical institutions in the form of electronic medical records is strictly for treatment purposes, and is not in a format intended for AI to learn from. In addition, current issues include the issue of data collection from the perspective of personal information protection, and the large amount of costs incurred in building systems and storage to utilize it. In order to promote AI drug discovery, it is necessary to overcome these problems and establish a system that can secure sufficient data.
Additionally, there is the issue of the length of time it takes to collect data. For example, there is a high demand for drugs to treat Alzheimer’s disease, the cause of dementia, but because the disease often only develops in old age, this disease was identified in order to gather basic comparative data. It could take decades to track a group of people over a period of time.
Movements by companies, universities, etc. regarding the use of AI in drug discovery
Companies and universities are also moving to promote AI drug discovery. In Japan, the Life Intelligence Consortium (LINC), an industry-academia collaboration project, was launched in 2016, and more than 120 organizations, including research organizations such as Kyoto University and RIKEN, pharmaceutical companies, and IT-related companies, are participating to discover approximately 30 types of drugs. We are proceeding with the development of AI.
In addition, the development of therapeutic drugs and preventive vaccines using AI has already begun for the new coronavirus infection (COVID-19), which caused chaos around the world in 2020. We are developing a technology that uses this to select multiple amino acid sequences that can serve as vaccine targets.
In this way, the movement surrounding AI drug discovery is likely to become even more active in the future.
By utilizing AI in drug discovery, it is possible to significantly shorten the development period. Additionally, because it can reduce development costs, AI drug discovery is attracting a lot of attention as a new method for developing new drugs. On the other hand, the challenge for the future is to create an environment that can collect the learning data needed by AI.