A multidisciplinary team of researchers from the Hangzhou-based Zhejiang University have made a major breakthrough in membrane proteins that could pave the way for tackling genetic afflictions like Parkinson's disease.

The study, published in Nature last month, shows the researchers have successfully engineered a set of artificial proteins that are able to regulate the functions of G protein-coupled receptors, or GPCRs.

GPCRs are the largest family of proteins located in the cell membrane. By binding extracellular substances, they play a critical role in transmitting a variety of extracellular signals from those substances into the cells and regulate diverse physiological functions.

More than 30 percent of approved drugs worldwide target these receptors, said Zhang Yan, vice-dean of Zhejiang University's School of Medicine and one of the leading researchers of the study.

"Currently, the vast majority of drug designs targeting GPCRs center on a natural site called the 'orthosteric site', or primary binding pocket, on these proteins," said Zhang. "The pocket can be considered the receptor's 'on switch', through which drugs exert their effects by either pressing or releasing the switch."

Abnormal genetic expression and mutations in the receptors can impair these switches and therefore disrupt their signaling functions. In fact, hundreds of clinical diseases, including Parkinson's, obesity and hypercalcemia (high level of calcium in the blood), are found to be caused by such mutations.

For patients, these structural dysfunctions often translate into long-term and chronic burdens, as conventional drugs designed to target the receptors' switches are generally unable to execute repairs.

Instead of focusing on these "switches", Zhang and his team focused on the structures of the receptors. Their goal was to design artificial transmembrane proteins known as modulators — a customizable, programmable "armor" or exoskeleton — and attach them to the key structure of the malfunctioned receptors, thereby enabling a more precise regulation of their functions.

"It is much akin to installing prosthetic limbs for persons with disabilities, or implanting medical devices supported by brain-computer interface technologies, only this time at the molecular level,"Zhang said.

Unlike conventional drugs, such an approach suggests a much more promising treatment avenue for disorders that result from genetic mutations in the receptors, as its therapeutic effects could be much more robust and last much longer.

In the study, Zhang and his team selected the dopamine D1 receptor, or D1R, as a prototypical model, and successfully made four modulators that could bind to the receptor and restore the activities of its various loss-of-function mutants.

But the process is far from simple and easy. To begin with, the design of the modulator itself poses daunting challenges. "For example, if a modulator is made up of 60 amino acids, and with 20 different types of amino acids available, the possible combinations would run up to 20 to the power of 60," Zhang said. "You can do the math."

That is where artificial intelligence comes to the researchers' aid, as the development of AI-driven protein design in recent years, particularly generative models for de novo design, has provided tools that generate proteins with unprecedented speed and accuracy.

The goal of de novo design — making new proteins from scratch — is to design proteins that do not exist in nature, without relying on natural structures or sequences, according to Zhang Min, another member of the research team from Zhejiang University's College of Computer Science and Technology.

Compared to modifying or improving existing proteins, this task presents greater challenges for researchers and algorithms in understanding proteins.

"We need to systematically deconstruct a seemingly simple requirement and transform it into functional modules that AI can implement step by step," she said.

However, the real hurdle lies in creating the right modulators to bind the receptors at the right places and achieve the desired effects.

To address that challenge, the team developed an AI-guided probe, through which the structures of the targeted receptors can be thoroughly profiled and potential binding and regulatory sites found. Then, using an approach known as "structural prompts" — not dissimilar to input fed to Deep-Seek or ChatGPT, but for protein structures — they generated these modulators.

"Not only have we designed modules that can 'switch off' and 'switch on' receptors or produce the intended signaling, we've even made them programmable to a certain degree," said Zhang, the computer scientist.

More significantly for the research team, their findings can now serve as a platform for similar research.

chenye@chinadaily.com.cn