“We have developed the world’s first machine learning solution that predicts a new drug’s clinical success rate with 70 percent accuracy.”

Deloitte’s Life Science & Health Care Group made such bold announcement at a news conference on Wednesday. The company said it has developed “D. Predict,” a new drug value evaluation program using artificial intelligence and machine learning, which was the first in the world.

According to Deloitte, the solution forecasts a success rate of a clinical study utilizing six machine learning algorithms, including deep learning. The company’s program learned from 13,000 clinical trials, including those registered in the U.S. Food and Drug Administration. Through the process, the solution showed more than 73 percent accuracy of predicting the success of the clinical trials.

“The possibility of clinical success in new drug development is a key factor in determining the market value of biotech shares,” said Oh Bong-geun, a partner at Deloitte.

D. Predict considered more than three times as many variables as the conventional method and used multiple machine learning algorithms such as deep learning to raise the predictability. Its prediction can guarantee more than 70 percent accuracy., he added.

According to Deloitte, D. Predict foretells a clinical success rate using 14 input data. The 14 are a clinical trial phase, indication, disease group, the gender of participants, the health status of participants, the number of participants, referred subject, research type, period, geographical location, drug type, drug’s mechanism, target, and route of administration. The results are divided into four categories -- successful, uncertain, unsuccessful, and partially successful.

D. Predict is a Web-based service, provided in a package with Deloitte’s consulting service.

However, the credibility of Deloitte’s prediction of new drug’s value remains questionable. No other credible institution has verified the program’s estimated results, except for Deloitte.

Deloitte is not carrying out any research or study regarding D. Predict. However, the company is applying for a patent.

It is also questionable that the prediction of the past clinical trials’ success could be applied to upcoming biological medicines.

D. Predict’s 73 percent accuracy in predicting clinical success came from learning of 13,000 clinical studies. Most of the past clinical studies were involved in synthetic chemical medicines, and biological drugs were very few. Deloitte acknowledged such argument, too.

New drug pipelines being developed by local drugmakers include many biological drugs, however. Local biotech ventures are also working on biological agents as novel drug candidates.

Deloitte said that D. Predict can show results from separate studies of biological drugs and that the company would add biological drug-related studies to overcome shortcomings.

Whether a success rate prediction of biological drugs alone will show 70 percent accuracy has not been verified yet. The company has not revealed the number of clinical trials associated with biological drugs, either.

Deloitte emphasized that D. Predict would help offset uncertainty about biotech shares and attract more investment in pharmaceutical firms and biotech ventures.

“From the investors’ point of view, new drug development takes a long time, and many variables occur, leading to irrational investment decisions. If the valuation becomes more accurate, though, investors will be able to make reasonable decisions, and more investment will flow into the pharma industry,” Oh said.

Deloitte is the world’s largest consulting service firm with 45,000 experts providing consulting for strategy and operation, technology, finance, and management. In Korea, more than 230 consultants are working under Deloitte. The company is also operating LSHC Practice Group for life science and healthcare sectors.

pkt77@docdocdoc.co.kr

<© Korea Biomedical Review, All rights reserved.>