Artificial Intelligence Leads to Efficient Drug Development

As artificial intelligence continues to develop in complexity, researchers have found new ways to harness that complexity in order to enhance research into treatments for diseases and other chronic conditions which have historically been difficult to treat. These developments indicate the potential for a change in the way in which drugs are developed and tested in the future.

The particular issue being looked at in this case is the sheer volume of data available to drug researchers. In spite of the amount of data available, the rate of successful drug discovery is lower than it was nearly 40 years ago. Put simply, drugs that are successful in preclinical trials tend to fail during clinical tests. The reason for this issue is simple: preclinical trials rely on things like lab mice, which are very similar genetically. In contrast, participants in clinical trials have far more genetic variety, which means that a treatment which works well in preclinical trials still has a significant chance of failing clinical trials.

A recently-published study at the University of California San Diego School of Medicine outlined a different approach to preclinical trials which use machine learning to better-predict whether or not a drug will make it through FDA trials. The methodology involves using artificial intelligence to model the gene expression of the disease during the onset and throughout its progression - meaning that it allows the system to more accurately predict how treatment will perform on a much wider variety of patients. In essence, the system can simulate the use of a particular type of treatment on a wide variety of genetic profiles using existing data rather than needing to generate new data.

This approach is called the Phase 0 approach, and it has the potential to save drug manufacturers millions of dollars. Rather than starting with preclinical trials, the effects of treatment can be predicted by using a model of the disease generated through machine learning. This gives a much better view of how a treatment or drug will perform in clinical trials. This in turn means that if a drug proves less effective when run through the model, manufacturers can avoid sinking money into doing preclinical and clinical trials. That money can instead go into potential treatments with better chances of being effective for more patients.

The results of the Phase 0 study still need to be developed through clinical trials to verify the effectiveness, but should the prediction of success prove true, it could revolutionize the way drugs are developed, particularly for chronic conditions.

More information on the study and methodology can be found here