We are already accustomed that a part of work in healthcare can be done by machines. They have proved to be effective in helping with diagnostics or treatment line selection, as well as with the document flow management. Yet, once we cast a glance beyond patient-healthcare providers relations, we are amazed at the scale of the problems and the potential for AI to solve them.
$ 2.5b and 10 years of research — these are the figures that describe the drug development process. Add up to them that only 1 in 10 drug would pass all necessary stages and eventually reach the patient. The present-day fast and furious world can afford neither such expenses nor such time frame.
It is here that AI techniques can add the most value, making the drug discovery quicker, cheaper and more effective. Some pharmacists are still skeptical, but most experts expect these tools to become increasingly important. If the proponents of these techniques are right, AI and machine learning will usher in an era of quicker, cheaper and more-effective drug discovery
For instance, McKinsey estimates that better decision-making, optimized innovation, improved efficiency of research, clinical trials, and new tool creation with the help of big data and machine learning could generate up to $100b in pharma and medicine annually.
AI has the potential to change the whole process of drug discovery. So far, the stages of drug development starting from a hypothesis and going toward testing the drugs are not connected at all. On the contrary, from a machine learning perspective, the stages become interconnected since you can use the data from the next stage to understand what happens in the previous stage or two stages before. Besides, the simultaneous access to multiple data can identify a quantifiable segment instead of using broad descriptors, such as disease symptoms. With machine learning, researchers can perform a trial on a pool of patients, receive different results and map them onto the patients’ genetics of molecular signatures, defining disease on a firmer ground.
AI has already been used successfully in all main stages in drug development:
· Stage 0. Literature overview
· Stage 1: Identifying targets for intervention
· Stage 2: Discovering drug candidates
· Stage 3: Speeding up clinical trials
· Stage 4: Finding Biomarkers for diagnosing the disease
Stage 0 Literature overview
There is an enormous amount of research that gets published every day and if we could collate the insights from all studies, we can formulate a better hypothesis. However, it is impossible for a human to read all abstracts and scientific papers, so researchers that work in the scientific domain usually just focus on one area and do not read other journals. But these journals contain a lot of relevant data that can inform decisions in the areas that a person is researching on. The solution is to let machines read all available literature, patents, and documents and pool the data together in a database of facts that can be extracted from this literature. That forms the basis of the hypothesis to find therapeutic targets for diseases.
Stage 1: Identification of targets for intervention
The first step in drug development is to understand the biological origin of a disease and its resistance mechanisms. To treat a disease, it is crucial to identify good targets, usually, proteins. The broad application of high-throughput techniques, such as short hairpin RNA (shRNA) screening and deep sequencing, has boosted the amount of data available for discovering viable target pathways. However, it’s still a challenge to integrate the high number and variety of data sources — and then find the relevant patterns. Machine learning algorithms are known to be good in such tasks and can handle all available data to automatically predict good target proteins.
Stage 2: Drug candidates discovery
With targets identified, researchers start looking for a compound that can interact with the identified target molecule in the desired way. This involves screening thousands and millions of potential natural, synthetic and bioengineered compounds for their effect on the target and their side-effects. Machine Learning algorithms can predict the suitability of a molecule based on structural fingerprints and molecular descriptors, blaze through millions of potential molecules and filter them down to the best options with minimal side effects.
Stage 3: Faster clinical trials
The key to successful trials is an accurate selection of suitable candidates, because choosing wrong entails prolongation of trials and waste of time and resources. Machine Learning can speed up the design of clinical trials by automatically identifying suitable candidates and ensuring that the trial participants are distributed among groups correctly. ML algorithms can identify patterns that would predict good candidates. Besides, they can notify the researchers that a clinical trial is not producing conclusive results so that the researchers could intervene earlier, and potentially save the development of the drug.
Stage 4: Identification of biomarkers for diagnosing the disease
Finally, you can only treat patients for a disease once you’re sure of your diagnosis. Biomarkers are molecules found in bodily fluids such as blood that provide absolute certainty as to whether or not a patient has a disease. They make the process of diagnosing a disease secure and cheap. They can be also used to pinpoint the progression of the disease — making it easier for doctors to choose the correct treatment and monitor whether the drug is working.
Yet, discovering biomarkers involves screening tens of thousands of potential molecule candidates. Once again, AI can automate and speed up the process. The algorithms classify molecules into good and bad candidates — and researcher can focus on analyzing only the best prospects.
Biomarkers can identify:
· Diagnostic biomarker: The presence of a disease as early as possible
· Risk biomarker: The risk of a patient developing the disease
· Prognostic biomarker: The likely progress of a disease
· Predictive biomarker: Whether a patient will respond to a drug
Even though broad application of AI is still in its infantry, there are multiple examples of its use by pharmaceutical companies. For example, the pharmaceutical giant Merck & Co is working on a project which uses deep learning technology for the discovery of novel small molecules. Pfizer has started a collaboration with IBM Watson for immuno-oncology drug discovery research. Researchers at the Massachusetts-based biotechnology company Berg have developed a model to identify previously unknown cancer mechanisms using tests on more than 1,000 cancerous and healthy human cell samples.
This shift presents suggests the industry has not only woken up to but is actively embracing the benefits of machine learning to identify and screen drugs, more accurately predict drug candidates and, ultimately, cut R&D costs and effort.
How will AI change the future of human experts?
As more studies are published and discussions are held around the future of AI in medicine, distinct sides to the argument emerge. The general consensus is that while routine tasks and data collection/entry should be done by machines, there will always be a need for the human element of the curator’s role, in judgment, creativity, and empathy or other human factors that modern technology cannot provide.
As curators, humans will set up the problem and let the algorithms or robots solve it. They will tailor and target specific compounds, symptoms, diseases or others rather than random or minute problems just for the sake of doing so. Besides, human experts will provide approval through the different phases of testing or exploring further options based on results based on context that the bots may not understand.
In conclusion, the future lies in cooperation between humans and machines and human clinical experts will have to adapt, learn and grow alongside technological advancements. Though future specialists will need to be both medical and computer experts, for medicine, it is evolution, not extinction.
