In this article, we talked about machine learning and Artificial Intelligence in pharmacy. It is a very wide subject, so we decided to draw more attention to one of the most important parts of it – Artificial Intelligence and machine learning in drug discovery. But this time we will go a bit different way. First, we will examine the traditional drug development process and then switch to AI drug discovery.
In general, drug discovery is a very time-consuming and costly process. It is estimated that putting on sale just one new drug, you have to invest at least 350 million USD*. And it takes around 10 years to do so. Machine learning in drug discovery may shorten and cheapen this process. That is why currently Artificial Intelligence in drug discovery gets more and more attention. Because at stake are hundreds of millions of dollars. But before that, let’s focus on the current situation. What does it look like to develop a new drug with the usage of traditional methods?
Traditional drug discovery
A short history
Have you ever wondered when the first drug was invented? If we think about a drug as an analgesic, then the answer should be 3400 B.C. At that time Egyptians, Babylonians, and Assyrians were using opium as a painkiller. The first drugs were 100% natural, made mostly out of plants and herbs. The next very important step in drug discovery was around the sixteenth century when a scientist called Paracelsus discovered that opium tends to work better with alcohol instead of water. Not much longer his invention was introduced to the market as Laudanum – a mixture made of cinnamon, opium, sherry, saffron, and clove.
Another very important milestone in the history of modern medicine was the invention in 1899 of acetylsalicylic acid (ASA), known as aspirin – nonsteroidal anti-inflammatory drug. Did you know that aspirin was hugely beneficial during the time of Spanish flu in 1918-1919? Even now it is commonly used, despite two main competitors on the market – ibuprofen, and acetaminophen known as paracetamol.
Artificial Intelligence in Drug Discovery – Where to start?
We will track the drug development process based on procedures provided by the FDA* – US drug agency. The first step in drug development is to understand the biological origin of a given disease and its resistance mechanisms. That disease is called “target”. Candidates for a drug are tested for their interaction with the target. Each candidate can be made of up to 10,000 molecules. Those molecules are subjected to a stringent screening process. When the interaction with the drug target is confirmed, scientists validate target by checking for activity vs. the disease condition for which the drug is being developed. In the end, one or more lead compounds are chosen. That’s because only a small number of compounds look promising for further studies.
Once we have chosen one lead compound it has to be thoroughly described. Size, shape, preferred conditions for maintaining its functionality, toxicity, bioactivity, and bioavailability – all of that has to be checked, determined and established. The next step is to describe its influence on the human body. Scientists gather information about: preferred dosage, potential benefits, side effects, interactions with other drugs, effectiveness compared to similar products and affect various groups of people (gender, race, ethnicity).
Artificial Intelligence in Drug Discovery – Research
Before conducting a test on a group of people, preclinical tests have to be done. The company must provide detailed information on dosing and toxicity levels of a newly developed drug. At this point, the most important aspect is to establish if the drug is safe to use. After preclinical testing, researchers review their results and decide whether the drug should have be proceeded to the clinical tests.
The next step is clinical research and trials on the group of human candidates. First, scientists have to design clinical trials to answer specific research questions related to a given medical product. Researchers have to decide:
· How many people will be involved in the trials?
· How long the study will take?
· Whether there will be a control group and other ways to limit the bias?
· What will be the dosage and how it will be administered?
· What clinical data will be collected?
· How will it be reviewed and analyzed?
When all of that is established, we can go to the next stage – clinical tests. They consist of 4 phases. Phase 1 takes several months and around 100 volunteers are involved. After this stage, 30% of drugs are rejected. Then, Phase 2 consists of several hundred people and lasts for up to 2 years. Phase 3 involves up to 3000 candidates who must have the given disease. Phase 4 is done with the participation of several thousand candidates. Through all these phases safety, side effects, efficiency, and adverse reactions are monitored.
You may also find it interesting – Data Science in Healthcare
Drug review and post-market monitoring
As we can read on the FDA website: If a drug developer has evidence from its early tests and preclinical and clinical research that a drug is safe and effective for its intended use, the company can file an application to market the drug. The FDA review team thoroughly examines all submitted data on the drug and makes a decision to approve or not to approve it.
In order to do that, the drug developer has to provide NDA – New Drug Application. NDA has to provide the following data: proposed labeling, safety information, drug abuse information, patent information, directions for use. If any studies have been conducted outside the United States, all of that data obtained from those studies has to be attached as well.
If NDA matches the FDA requirements, the next step is labeling. At this point, prescribing information has to be developed and refined. It happens in cooperation with FDA and drug developer company. This is the moment when the FDA can also ask additional questions to the provided data or demand additional tests.
Introduction of a new drug to the market
If the drug passes all tests and the provided data meets the FDA’s requirements, it can be introduced to the market. That doesn’t mean that all testing and monitoring is done. The last part is called “Post-Market Drug Safety Monitoring”. This is necessary because, as the FDA states, it is impossible to have complete information about the safety of a drug at the time of approval. So even for months and years after introducing the drug to the market, monitoring lasts.
During that monitoring, the FDA reviews reports of problems with prescription and conducts inspections of drug manufacturing facilities. What’s more, if drug developer intents to change already approved drugs in any way, it also has to be approved by FDA before the change is applied.
Traditional drug discovery – sum up
To recap, we have to admit that developing a drug is a costly and time-consuming process. It’s estimated that developing just one medicine costs at least 350 million dollars, and it is the lowest level. No one’s surprised with the amount exceeding 1 billion dollars! Did you know that whopping
95% of the experimental medicines that were successfully passed to clinical tests had been inefficient and unsafe? The complexity of this process is the very reason why it’s so costly.
And the effects are clearly visible – 66 of the 98 companies studied by Forbes launched only one drug this decade. Things get worse, because as the same studies show, the more drugs you develop, the higher is the cost for each one drug developed. Therefore, it hits a level of almost 5.5 billion dollars for companies that have brought to market at least 8 medicines over a decade.
All in all – pharmaceutical companies are constantly looking for ways to shorten and cheapen this process. At stake are big numbers but also human health. We do not need to say how many benefits it would bring if the drug discovery process was shorter and cheaper.
The sad reality is this – currently many rare diseases are still waiting for the cure to be invented because in many cases it is not profitable for the pharmaceutical companies to look for them. So patients are waiting. If it would not be the issue, the lives of millions of people might be totally different.
But for that to happen, we need faster and cheaper solutions. And they are available! Almost at hand! Machine learning and Artificial Intelligence in drug discovery. Let’s see how they can be game-changers in drug discovery.
Machine learning drug discovery – forecasts
Developing a drug with the usage of Artificial Intelligence and machine learning can change the way we think about developing a drug. Before we turn to details, let’s check the forecasts for the future. As we can read in the Bekryl Intelligence report “Artificial intelligence has the potential to offer over US$ 70 billion saving for the drug discovery process by 2028. The potential to boost a company’s ROI along with its time-saving process has led big pharmaceutical and biotech companies to invest heavily in technologies.”
If we assume, that single drug discovery costs around 350 million dollars, then at stake are great savings. It’s definitely worth the efforts. So the direction is clear – we have to invest in Artificial Intelligence in drug discovery and machine learning in drug discovery if we want to see more new drugs on the market. This is the best example of how new technology can be beneficial to society in a very measurable way.
Discover other Machine Learning Solutions.
Artificial Intelligence in Drug Discovery – how?
AI drug discovery can be beneficial in a number of levels, but the main are:
· designing a drug’s chemical structure
· investigating the effect of a drug – both in basic preclinical research and clinical trials.
Those are the two most important stages of the drug development process. And the most time-consuming as well. First, let’s examine the influence of AI drug discovery in designing a chemical structure of a drug.
Artificial Intelligence in Drug Discovery – Chemical structure
Artificial Intelligence truly has become a crucial component of drug discovery. As you already know, the first step in drug development is to understand the biological origin of a given disease. That requires huge amount of data to be processed. Machine learning in drug discovery can do that easily and analyze thousands of sources just in seconds. For instance, MIT scientists have created a system that analyzes available science data and on that basis makes a chemical compound. Researchers have tested the device and, with it, managed to make 15 different medicines. Almost without human assistance! For example, they have managed to create a machine-made aspirin, and four non-steroidal anti-inflammatory drugs.
Another company, called AtomNet, has made a deep learning neural network application made to help in drug discovery. The company behind the project says it is based on CNN technology (Convolutional Neural Networks) and helps by combining atoms to come up with possible molecules. The company says that its algorithms are “trained” to analyze billions of atoms to find which will bond together to finally create an efficient compound. This is very helpful because the traditional process to find the right compound takes a lot of time. Machine learning in drug discovery algorithms do that much quicker and the result is almost ready to proceed to the next step.
So machine learning in drug discovery can help to shorten the process of designing the proper chemical structure. This is huge saving of time because Artificial Intelligence algorithms can analyze the mind-boggling amounts of data in a much shorter time.
Artificial Intelligence in Drug Discovery – Research
This is another field, where Artificial Intelligence in drug discovery can be very beneficial to the industry. Researches are divided into two groups: preclinical testing and clinical tests. With the usage of a strategy of creating desirable compounds scientists achieve two goals: first, they can create a compound optimized for a specific characteristic, and second, they can reduce the unwanted side effects of a medicine. This is very helpful when the phase of clinical tests comes into play.
Machine learning in drug discovery can do almost all of the preclinical testing by itself. No human patients are involved at this stage, it all happens “virtually”. So all you need to conduct successful preclinical tests and obtain an insightful report that summarizes tests is data to base on and a way to analyze it in an efficient way. Artificial Intelligence is simply designed to execute exactly those tasks.
As you know, every new compound has to be thoroughly described. Scientists are establishing its size, shape, preferred conditions for maintaining its functionality, toxicity, bioactivity, and bioavailability – all that data can be easily gathered with the assistance of machine learning in drug discovery. When the compound is made and properly described, the next step is clinical tests.
Artificial Intelligence in Drug Discovery – Clinical tests
Clinical trials are the most expensive stage of drug development, just to mention the need to pay every patient taking part in the tests. Simultaneously, they are the most important ones because they aim not only to verify given medicine’s efficiency but also safety. This is why pharmaceutical companies pay a lot of attention to this phase.
One of the key factors to conduct successful clinical tests is to find suitable candidates. Either healthy and with a given disease. Representing many (preferably all) races, ethnicities, ages, and genders. And what’s more, all of the patients taking part in the tests have to be thoroughly examined before they are allowed to participate in the testing. AI drug discovery can considerably speed up the process of looking for these candidates and rejecting unsuitable ones. For example, AI drug discovery analytics can analyze genetic information to identify the proper patients for the tests. Machine learning drug discovery algorithms can also reduce data errors. The result? Faster and more accurate clinical tests.
AI drug discovery – the biggest milestone?
The time has come to summarize our considerations. We’ve established that developing a drug in a traditional way is a tricky thing. Just one out of 10 medicines eventually reach the market. Many medicines are rejected due to their ineffectiveness or danger to human health. And above all that, pharmaceutical companies have to invest literally hundreds of millions of dollars to develop a single drug.
Machine learning and Artificial Intelligence in drug development can be beneficial in many ways. Starting from designing the desired chemical compounds, that are a basis for future drugs, through analyzing huge amounts of data in order to minimize unwanted side effects up to the gathering proper candidates for the clinical tests. Artificial Intelligence in drug discovery is a great milestone that assists human scientists in every single step of the long and winding journey to the discovery of new medicine.
Cheaper drug discovery means more drugs on the market
Because it is faster and more accurate, the whole process is simply much cheaper. That leads to very optimistic forecasts – cheaper drug discovery means more drugs on the market. Which is heartwarming for the patients with rare and difficult to cure diseases, for which current pharmaceutical offer is simply insufficient. Who knows, maybe in not too distant future, drug development will take just a year instead of the current 10 years? And the drugs will be much more effective and… much cheaper? This is the real possibility, thanks to the machine learning in drug discovery and AI drug discovery.
Do you run a pharmaceutical company? Are you interested in the implementation of Artificial Intelligence in your company? Just give us a call! We are always vitally interested in the cooperation with companies wanting to start a new phase of their development – with machine learning and Artificial Intelligence.