With the average cost of bringing a new drug to market now at $2.6 billion1 and one-tenth of drug candidates failing to make it to market despite successfully completing Phase I trials2, it's no surprise that pharmaceutical companies have seized on the unparalleled data-processing potential of artificial intelligence (AI) systems.
Their use in discovering chemicals, some of which may have previously finished clinical trials, that may be quickly and cheaply repurposed to treat other disorders is widely documented. However, as study experts are discovering, AI systems are capable of far more.
In an era when speed and efficacy are critical, the worldwide artificial intelligence (AI) in drug development market is undergoing seismic shifts. Based on the emulation of human intellect by machines, this cutting-edge technology is assisting in the identification of new compounds, the development of tailored treatments, and the streamlining of difficult challenges in the pharmaceutical business. Machine learning and deep learning, two critical components of AI, are being used by companies to fundamentally improve their drug discovery processes. By shortening timescales, AI-based firms may identify, design, and optimize novel treatments in less time, often exceeding traditional approaches.
Traditional drug development is primarily reliant on human-derived logic and effort to find disease functioning mechanisms, identify druggable targets, and design lead compounds to hit the targets. Despite breakthroughs in our understanding of human diseases and biotechnology, the search for innovative treatments remains a time-consuming and expensive procedure. With the recent phenomenal success of artificial intelligence (AI) in a variety of disciplines, AI-based drug development is poised to become a revolutionary force in the pharmaceutical sector, radically changing the old trial-and-error design method.
Why drug developers need AI and benefits
The potential applications of AI in drug discovery are nearly limitless, although repurposing existing medications has been one of the main areas of research to date. This typically entails discovering new applications for medications that have previously received market and regulatory approval for the treatment of a specific ailment. It is feasible to identify whether a medicine molecule will bind to other specific targets by using AI technology to analyze current research data, which may include information from clinical trials and other patient data. This information can then be utilized to determine how effective the medicine may be at other dosages or when treating new patient populations. When employed in this manner, AI systems can assist in identifying re-purposing opportunities more rapidly and efficiently than traditional scientific research approaches.
Machine Learning for Pharmaceutical Discovery and Synthesis (MLPDS) is an MIT Consortium that brings together computer scientists, chemical engineers, and chemists from MIT with scientists from member companies to develop new data science and artificial intelligence algorithms, as well as tools to aid in the discovery and synthesis of new therapeutics. MLPDS trains scientists and engineers to work effectively at the data science/chemistry interface and allows member companies and MIT to collaborate to create, discuss, and evaluate new data science advances for chemical and pharmaceutical discovery, development, and manufacturing.
Synthesis planning; prediction of reaction outcomes, conditions, and impurities; prediction of molecular properties; molecular representation, generation, and optimization (de novo design); and extraction and organization of chemical information are among the specific research topics addressed by the consortium. The algorithms are created and confirmed using public data before being distributed to member companies for use with proprietary data. All members have equal access to intellectual property and unfettered access to all advancements. Through one-on-one meetings and teleconferences with individual member organizations, Microsoft Teams channels, GitLab software repositories, and consortium face-to-face meetings and teleconferences, MIT strives to make tool creation and transfer successful.
Indeed, AI systems have the ability to provide a more definitive picture of the potential of a repurposed medicine than would otherwise be achievable. AI systems can identify whether a pharmacological compound binds to many targets and whether the drug has the ability to treat diseases associated with one or more of the targets by binding to such targets. Unlike human research, the analysis delivered by AI systems is guaranteed to be objective because it is based on patterns drawn from recognized data sources.
The impact of AI on drug development can be considered revolutionary, primarily because of its ability to significantly reduce the usual drug design timeframes and expand the current drug discovery scale. Here are four key benefits of AI in drug development:
1. Objectivity
AI is a scientific discipline. It means that the AI-based drug development process is not influenced by prejudice, prior information, personal interests, or anything else that could have a direct impact on the development outcomes.
2. Constant advancement
AI, in addition to assisting people with routine jobs, makes use of cutting-edge biological and computing technology. This process is always growing, and with increased powers of invention and lower costs of AI technologies, AI plays an important part in drug development developments now and in the future.
3. Higher predictivity
The greater predictability capacity of AI techniques in drug screening has a positive impact on the process of establishing meaningful interactions. As a result, a diligent approach to parameter design can dramatically reduce false positives.
4. Streamlined process and reduced human inefficiencies
Drug screening outputs can be optimized with the use of a virtual lab, drastically lowering human resource hours.
Use Cases of AI in Drug Discovery
The importance of AI in medication development cannot be overstated. One of the key requirements for measuring the impact is that AI can be used across the entire lifecycle of pharmaceutical products.
Use case 1. Toxicity prediction
It is impossible to prevent a medicine's hazardous effects without first forecasting the toxicity of the drug molecule. One of the examples of such kind of predictions is the DeepTox machine learning algorithm which was able to:
detect static and dynamic features in the chemical descriptors of the molecules
predict molecule toxicity based on over 2000 predefined toxicophore features
Use case 2. Target protein structure prediction
Several proteins have a direct impact on disease progression. As a result, predicting the structure of the target protein is critical while developing the therapeutic molecule. Because AI can predict 3D protein structure, it may be useful in structure-based medication development. The AlphaFold AI tool is a nice example of such AI tool leveraging. AlphaFold was able to:
analyze the distance between amino acids and the corresponding angles of the peptide bonds
predict the 3D structure of the target protein
predict correctly over half of 43 3D structures.
Use case 3. Pharmaceutical manufacturing
AI tools are widely used in drug manufacturing. E.g. DEM AI tool streamlines the manufacturing process by:
predicting the perspective tablet path during the coating
analyzing the time that is spent by tablets under the spray zone
analyzing varying blade shapes and speed
studying the segregation of powders
Use case 4. Quality control and quality assurance
Several AI tools are utilized to control in-line manufacturing operations. To do this, the artificial neural network employs multiple methods, including self-adaptive evolution, backpropagation, and local search:
predict the desiccated-cake thickness and the temperature in the future time period in predefined conditions, and thus
maintain the final product quality check.
Challenges of AI in Medicine
Of course, the issue with this use of AI systems is that the quality of the datasets used determines the quality of the analysis offered. As a result, the pharmaceutical industry is increasingly seeking to collaborate, pool data, and use it to train algorithms through a machine learning process. The Melloddy Project, a recent project comprising ten pharmaceutical companies, including GSK, Johnson & Johnson, and AstraZeneca, is adopting a revolutionary blockchain method to store data on a secure ledger while protecting individual companies' trade secrets.
The increasing use of AI in finding re-purposing prospects may aid in the identification of therapies for additional diseases that affect smaller subsets of the population or persons in third-world nations where funding for drug discovery programs is scarce.
There are numerous examples of early success with repurposed medications. Deep neural networks, designed by AI specialist Atomwise, are being used to improve drug development by analyzing simulations of molecules in order to reduce the time that research scientists must spend synthesising and testing drugs. In 2015, the company used its AtomNet technology in collaboration with IBM and the University of Toronto to analyze and predict the molecules that could potentially bind to a specific glycoprotein in order to find a treatment for Ebola virus infections that had killed over 11,000 people in Africa and other parts of the world. Merck has recently used AtomNet to scan its existing medicines for chances to repurpose them to fight present or prospective diseases.
Despite AI evangelists' profound assertions about AI and machine learning bringing value to human employment, the intense argument over whether AI can replace human workers never seems to stop. Another prominent topic is the black box phenomenon, which refers to the lack of visibility of the processes and workings that occur between inputs and outputs.
The difficulties of data management and a shortage of skill can only be handled if a business recognizes the clear benefits of AI adoption and is willing to set the correct priorities to achieve maximum operational efficiency. Concerning the black box phenomena, explainable AI, i.e. a system in which the entire process of AI workings is straightforward and easy to grasp for humans, can help.
The Future of AI in Medicine
It is believed that AI technology will provide improvements in areas of drug discovery that receive less funding. Big pharmaceutical corporations are expected to remain focused on finding a cure for the most common and debilitating diseases, such as cancer and Alzheimer's. The increased use of AI in finding re-purposing prospects could aid in the identification of solutions for additional diseases that affect smaller subsets of the population or persons in third-world nations where financing for drug discovery programs is scarce. Pharnext is an example of a firm that uses machine learning to uncover compounds for the treatment of rare conditions, and it presently has a molecule in clinical trials for the treatment of Charcot-Marie-Tooth disease, a rare neurological syndrome. Because the original chemical has previously been proven safe, the repurposed form can reach the market faster because some portions of the clinical trials can be skipped.
While some see AI as a priority, most large corporations are not pushing for greater AI at this time. While AI's accuracy has advanced to the point where it can identify drug development targets, there are significant issues when dealing with personal patient information. According to a survey of significant US health systems, health systems prioritize strong cyber security over AI research at the moment.
Some firms intending to use more AI are still seeing poor progress. The difficulty here is finding new specialists who can fill the new positions of AI computer experts. Because AIs must be programmed with real data, this is also expected to create a huge number of new jobs at a faster rate than employment loss.
In a nutshell, AI has the potential to make a big technological contribution that will transform the entire drug discovery process. However, as many believe, AI-based successes, which are currently regarded as breakthroughs, will soon become common practice in the pharmaceutical sector. A defined and step-by-step AI adoption strategy can help life science organizations expand their horizons and win in a competitive market.
Comments