AI Has the Potential to Revolutionize Drug Discovery If Experts Are Involved
Mon, April 19, 2021

AI Has the Potential to Revolutionize Drug Discovery If Experts Are Involved

AI has the potential to revolutionize drug discovery if it used to augment scientists' abilities / Photo Credit: Sisacorn (via Shutterstock)

 

50% of late-stage clinical trials fail because of ineffective drug targets, which means only 15% of drugs advance from Phase 2 to approval, as cited by Alix Lacoste of health-oriented news website Stat. Hence, AI can help expand the drug  discovery field “by making more predictions in more novel areas of biology and chemistry.” AI can identify relevant information and create links between biomedical entities with little information. This can be done by extracting text from scientific research papers. 

In the last two decades, 50 trials have failed to present any positive results. Sadly, only two approved drugs on the market have only shown modest benefits to patients. This evidently calls out for new approaches. AI has the potential to identify new targets for disease quicker. It is also low-cost and has low failure rates. However, adopting AI in the medical field is still low due to trust issues. 

Sometimes, AI makes mistakes. For example, it may experience difficulties in distinguishing potential positive biological effects from the potential negative effects on the disease course. Further, AI may predict drug targets that scientists know will have side effects. Therefore, doctors need to help it by instructing to “filter out specific drug or target classes.” If successful, the AI system is guaranteed to produce the best scientific results. 

The role of AI in designing new drugs, repurposing old ones, or identifying targets should be used not to replace scientists’ abilities, but to augment them. Scientists are responsible for “determining the data to use” in the AI system and providing expert evaluation of the results for accuracy and nuance. 

Further, AI can be used to form hypotheses that might appear unlikely at first glance. Traditional drug discovery can only evaluate a finite set of evidence or experiments at one time, thereby increasing the likelihood of bias. Since scientists only see a part of the picture, they end up basing their assessment on the data and materials they have chosen for review. Thus, AI can help scientists avoid implicit bias that arises when only limited, local data are used to form a conclusion.  

According to Lacoste, scientists must advance their understanding of AI, including its uses, in drug discovery.