The 2017 Korea Bioplus conference held its main sessions on new trends in bio-drugs later on Monday, keeping on track with the conference’s goal on diagnosing global biotechnology issues.

The 2017 Korea Bioplus, which marks its third year, was attended by Korean and foreign institutions and businesses related to biotechnology, including the Massachusetts Institute of Technology. It focused on two primary topics – the Fourth Industrial Revolution and biotechnology, and bio-ecosystem for startup.

The sessions focused on the Fourth Industrial Revolution and biotechnology and went into detail on AI’s role and capability in the biotechnology field.

The first session, led by Ji Xin, CEO of Shanggong Medical Technology, dealt with the company’s Autoeye AI system.

Autoeye uses RetinalNet, a deep learning framework named Caffe based on a CNN model, to collect 120,000 images from over 300 hospitals in China. It is used to diagnose diseases such as diabetic retinopathy and glaucoma. The system has 300,000 fundus images and is likely to hold over 3 million fundus images by 2018.

“RetinalEye is important as China has a low ratio of doctors per patient,” Xin said. “RetinalEye can help make an accurate diagnosis and speed up the process for doctors.”

Although the AI provides a diagnosis it is not absolute, and doctors must verify the results, he cautioned. The company has also started to support big data platform for new drug development with its vast advantages in patient recruitment covering 20 provinces in China. The broad coverage is expected to facilitate clinical trials of new drugs.

“The company expects in the future to expand its service pattern to cover cardiovascular and cerebrovascular diseases,” Jin said. “We will also launch a patient management service using continuous glucose monitors and obtained blood glucose data from diabetic patients.”

Kim Jin-han 김진한, CEO of Standigm, led a session in discovering novel drug candidate substrates with AI.

Standigm uses StanAI to discover or reposition new drugs and provide clinical supports such as patient stratification and matching.

StanAI performs in a variety of aspects such as providing interpretation on graph databases, which include the information about drug and disease or drug and target interaction. Its application also expands to filtering drug candidates by scoring candidates and their mechanisms and conducting an automated literature survey.

“Interpretable models are crucial in healthcare applications,” Kim said. “The AI has to consider which pathways or genes are responsible for the drug and disease relationship and look for other drugs that share the same pathways or genes along the path to the disease.“

The company’s continuous efforts have led to discovering various drug candidates in 2017. Their most promising finding is with Ajou University hospital, where they are currently validating their new drug substance for Parkinson’s disease through animal trials.

“Unlike mathematical models based on data, biology has no definite guideline or rules,” Kim said. “Even if AI finds a suitable drug substance the drug has to be proven effective through clinical trials.”

Drugs targeted for patients have to go through more rigorous trials, Kim said.

“Our company is not looking forward to solving every problem in the medical field using AI,” he added. “If we can reach areas that are too vast and deep for the human brain to acknowledge or even touch it will be a grand feat for AI applications in the biotechnology field.”

Copyright © KBR Unauthorized reproduction, redistribution prohibited