How does AI pharmaceutical manufacturing protect human health?

Introduction

In today's society, medicine is developing rapidly, more and more diseases are being conquered by humans, and the average human life expectancy is also constantly increasing.

As humans develop new drugs and expand their arsenal of weapons to fight diseases, AI is also contributing. Moreover, with the addition of AI, human research and development of new drugs may also enter a new era.

Difficulties in traditional drug development

In the field of drug research and development, there is a term called the "double ten dilemma" or the "double ten law."

It is said that it takes more than 10 years and more than 1 billion US dollars to develop a new drug from research to market. In addition, there is actually a hidden 10: the success rate of new drug research and development is only around 10%.

Why is it so difficult to develop new drugs? This starts with the development of new drugs. From the development to the launch of a drug, there are roughly the following stages:

First, the drug development stage:

This stage includes determining the target of the drug, that is, the location where the drug acts on the target cell or virus. Then, targeting the target, designing drug molecules and making candidate drugs.

Candidate drugs cannot be tested on humans immediately. They must first be tested in the preclinical stage in the laboratory to understand drug pharmacokinetics, pharmacodynamics, and preliminary safety evaluations.

Drugs that have passed the preclinical stage can enter the clinical research stage, which includes three phases:

Phase I mainly targets healthy volunteers to study the pharmacokinetics and safety in humans.

Phase II: preliminary evaluation of the therapeutic effect of the drug, focusing on observing efficacy and adverse reactions.

Phase III is a comprehensive evaluation clinical trial, which is the stage of confirming the therapeutic effect of the drug. The large-sample randomized double-blind experiment we are familiar with is in this stage.

After passing these three phases, the drug can be marketed. Of course, after the drug is marketed, there is still a post-marketing safety monitoring phase.

It takes about 12-15 years from the start of drug research and development to market launch, and in this process, a large number of drugs will be eliminated.

How AI can improve:

However, with the development of AI technology, drug research and development will enter a new era.

For example, when searching for drug targets, the selection and determination used to be based on the experience of human scientists and the reference of literature. However, the emergence of natural language processing technology and various large models has allowed AI to participate in this process. AI can not only read a large amount of literature and extract keywords, but also make inferences based on contextual semantics to assist doctors in finding potential targets.

In addition, after finding the target, it is necessary to crack the target structure in order to design a drug that can act on this target. AI can also provide assistance in this process.

According to data from Exscientia, a British AI pharmaceutical company, AI has been able to shorten the drug development phase from an average of 4.5 years to about 13.7 months, a reduction of nearly 75%.

During the COVID-19 pandemic, Pfizer used AI technology to accelerate the development of the special-effect drug Paxlovid, and even reduced the drug development process to 4 months.

In the preclinical stage, the safety and pharmacokinetic characteristics of candidate drugs can also be predicted based on AI models. Although AI cannot replace humans in conducting experiments, it can also help human scientists make some predictions and speed up efficiency.

Similarly, human experiments are essential in the clinical trial stage, but AI can also play a role. For example, in recruiting volunteers, AI can directly search electronic medical records to find the most suitable experimental subjects.

In addition to developing new drugs from 0 to 1, AI can also play an important role in "new uses for old drugs."

Drugs already on the market have been tested for safety and have detailed pharmacokinetic data, so if these drugs can be used to treat another disease, it will be faster and safer than developing a new drug.

But scientists cannot try aimlessly, and AI's natural language processing and deep learning technologies can help in this process.

Challenges and prospects of AI drug development:

It is worth noting that the AI ​​model can make accurate predictions about the efficacy and safety of drugs, which requires high-quality scientific research data. However, human scientists are still needed to evaluate and test the AI ​​prediction results.

But I believe that with the joint efforts of humans and AI technology, the pharmaceutical industry will undergo earth-shaking changes, and AI will promote the development of new quality productivity in the pharmaceutical industry.

Author: Yunjiyu Science Creation Team

Reviewer: Qin Zengchang, Associate Professor, School of Automation Science and Electrical Engineering, Beihang University

The article is produced by Science Popularization China-Creation Cultivation Program. Please indicate the source when reprinting.