How Can Artificial Intelligence Facilitate New Drug Research and Development?

Posted by Lisa George on November 27th, 2020

Owing to its powerful computing power and ability to discover relationships, artificial intelligence (AI) is applied in multiple procedures of new drug R&D. It can assist in discovering both explicit and hidden relationships between drugs, diseases, and genes, which would take a lot of work for drug experts if done manually.

In terms of computing, AI possesses powerful cognitive computing capabilities, which can perform virtual screening of candidate compounds, quickly screen out compounds with higher activity, and prepare for later clinical trials. According to statistics, AI is mainly used in 7 big scenes of new drug development, including target discovery, compound synthesis, compound screening, crystal form prediction, patient recruitment, optimization of clinical trial design, and drug redirection.

AI can save 40%-50% of the time in compound synthesis and screening compared with traditional methods, saving pharmaceutical companies US billion each year in terms of compound screening costs. In the clinical research phase, it can save 50% to 60% of the time and save 28 billion US dollars each year. In total, AI can save pharmaceutical companies almost US billion in R&D costs each year.

Drug target discovery

Traditionally, target research qualitatively speculates the relationship between the structure and activity of physiologically active substances in an intuitive way, and then discovers the targets on the body's cells where drugs can play a role. Pharmacologists refer to relevant scientific research literature and personal experience to speculate the target, which takes 2 to 3 years, and the possibility of finding the target is extremely low. AI uses natural language processing technology (NLP) to learn from massive medical literature and related data, and uses deep learning to discover the relationship between drugs and diseases so as to find effective targets within the shortest target discovery cycle.

Compound synthesis

Compound synthesis is a step to analyze the drug properties of small molecule compounds, including the ability to bind to the target, pharmacokinetics, pharmacokinetics, etc., to discover compounds with better drug activity and efficacy, and then design the synthesis according to a specific route. Pharmacologists and chemists will sequentially perform computer simulation tests on tens of millions of compounds. It will take several years to find compounds with better activity for synthesis, and the cost is usually tens of billions of dollars. AI uses its machine learning and deep learning capabilities to simulate the drug properties of small molecule compounds, and can select the best analog compounds for synthesis tests within a few weeks, and thus reduce the test cost of each compound to 0.01 cents.

Compound screening

The targeted proteins and receptors of each drug are not specific. If it acts on non-targeted proteins and receptors, it will cause side effects. For new drugs that have not yet entered the stage of animal testing and human testing, they need to be tested and judged in advance for their safety and side effects, and drugs with higher safety have been screened out. At present, high-throughput screening methods are mainly used for compound screening. At the same time, millions of tests are carried out by robots, and the screening cost should be tens of billions of dollars. AI can be applied into compound screening scenarios from two aspects. One is to use deep learning and computing capabilities to develop virtual screening technology to replace high-throughput screening, and the other is to use image recognition technology to optimize the high-throughput screening process. In this way, billion in compound screening costs can be saved each year.

Crystal Form Prediction

Different small molecule crystal forms have different drug stability and solubility. Therefore, the stable crystal structure is related to the quality of the drug. There are polymorphisms in small molecules, some of which have strong stability but poor solubility, and some have good solubility but poor stability. If you rely solely on manual labor to obtain a crystal form with strong stability and good solubility, it will not only take a lot of time, but the probability of success is also extremely low. The emergence of AI can greatly improve the effect of crystal form prediction. It relies on deep learning capabilities and cognitive computing capabilities to process a large amount of clinical trial data, and can find the crystal form with the best efficacy in a few hours or even minutes.

Patient recruitment

Before a new drug gains regulatory approval, three phases of clinical trials are required. Finding suitable patients is the prerequisite and basis for clinical trials to be carried out. Trial managers need to identify patients who are eligible for drug trials from a large number of cases. AI relies on deep learning capabilities to extract relevant information from massive clinical trial data, automatically match test results with patient conditions, improve the efficiency of precise matching, and complete trial recruitment and enrollment in a shorter time.

Optimize clinical trial design

The clinical research phase of drugs includes trial plan design, trial process management, trial data management statistical analysis, etc. If relying only on manual labor, it will not only be a lot of work, but also prone to errors. Likewise, AI's machine learning and cognitive computing capabilities can be widely used in processes like experimental research design, experimental process management, and statistical analysis of experimental data to improve the efficiency of the entire clinical trial.

Drug redirection

New use of old drugs is currently a common way to find drugs. It works by cross-studying and matching the drugs that have been exposed on the market and more than 10,000 targets on humans. Relying on AI's powerful natural language processing capabilities and deep learning capabilities, extracting knowledge that can promote drug research and development and new verifiable hypotheses from the disorganized mass of information will bring an exponential increase in the speed of trials. The application of AI in drug redirection can eliminate two steps: target discovery and pharmacological effect evaluation, which is expected to reduce the cost of drug development to 300 million US dollars or less, and the development cycle will also be shortened to 6.5 years.