The hunt to find the one in a billion molecule

The search for new brain drugs means sifting through more molecules than there are stars in the Milky Way galaxy

Not long ago, as recently as the 1990s, scientists searching for new drugs tested molecules one by one in the lab, looking for signs of disease-fighting potential and the next breakthrough medicine. These collections of molecules often ranged from a few hundred to a few million, representing naturally-occurring molecules as well as a growing ability to make molecules in a lab.

Testing these compounds ranged from very manual efforts in small labs to highly automated systems in specialist labs, called 'high throughput screening centres'. This is how many of our existing medicines were first discovered.

Today, with global advances in chemistry, the scale of these collections has massively increased, moving into the billions, even knocking on the door of trillions, of compounds.

“Just to put that in perspective, the search space for new medicines now contains many more molecules than there are stars in the Milky Way. What those numbers mean is that we have to rethink how we search for the molecules that could become our future medicines,” says Associate Professor Jack Flanagan, a pharmacologist from the University of Auckland’s Centre for Brain Research.

Flanagan leads a $1m Drug Discovery programme funded by the Neurological Foundation and supported by a team of scientists – Professor Mike Dragunow, Dr Daniel Conole and Dr Hamid Abbasi. The programme utilises advances in supercomputing and artificial intelligence (AI) to screen hundreds of millions, even billions, of molecules.

Known as ultra-largescale drug screening, it is changing how we tackle brain disease in New Zealand. “Such massive scale creates the compelling idea that hidden among those billions of molecules is one, the one molecule, that will successfully navigate the drug development process and actually do the job of getting into the brain and treating neuroinflammation,” Flanagan says.

Neuroinflammation is a process where cells of the immune system attack healthy brain cells, contributing to the development and progression of many brain disorders including Alzheimer’s, Parkinson’s, epilepsy, stroke and concussion.

It was chosen as the target for the programme as there is a huge lack of treatments for neuroinflammation, and it is implicated in so many brain diseases. The team is building on an existing ‘library’ of hundreds of millions of molecules screened with early support from the Hugh Green Foundation – who will help to co-fund this next phase of research.

Since the new drug discovery programme got underway in late 2025 it very quickly cracked the 700 million molecule mark. “Our target of one billion is already very, very close,” Flanagan says.

At the heart of the project is a technique known as molecular docking – using computers to predict how well a potential drug molecule might bind to, and turn off, a target protein involved in disease. “Molecular docking simulates the interactions between a potential drug molecule and a target protein using their three-dimensional structures, at the atom level. We then look for molecules that make the interactions required for an effective drug.”

The next planned stage is to test the most promising compounds directly on human brain cells and tissues. “These are the exact cells and tissues that a neuroinflammation drug needs to act on, so bringing them to the forefront of our discovery will accelerate our progress to real-world treatments.

This in-vitro (outside of a living organism) testing takes place at the Hugh Green Biobank at the Centre for Brain Research, which has unique expertise in growing cells from donated human brain tissue into living drug test systems.

Alongside the in-vitro testing, the project brings together a pipeline of specialists to bring treatments to patients as quickly, safely and effectively as possible, including AI modelling expertise from the Auckland Bioengineering Institute, led by Abbasi (with Research Assistant Shanan Chand), and clinical insight from doctors working directly with patients – all backed by the supercomputing power of the New Zealand eScience Infrastructure (NeSI).

“When you put it all together, this patient-centred drug discovery process is something that has not been seen before in New Zealand or worldwide,” Flanagan says.

“We’re not just theorising. We are combining global advances in chemistry, national supercomputing infrastructure, AI, real human brain tissue donated by New Zealanders, and clinical need, to move potential treatments closer to reality.

“It’s really important to us that New Zealanders are at the front of this pipeline, choosing the molecules that work for us. That is why the taonga of donated human tissue and the Hugh Green Biobank is so important in our work.”

By the end of the three-year programme, the team aims to take the most promising leads to pre-clinical trials and start the next phase of progressing treatments to patients. As part of that vision, the team wants to enable other researchers to use their model of patient-centred molecular discovery technology, enhancing the New Zealand neuroscience research landscape and moving knowledge of brain disease into future drug discovery.

Part of the inspiration for the programme came from what Flanagan and colleague Professor Mike Dragunow had seen taking place in developing cancer treatments.

“In oncology, we’ve been designing drugs to target specific enzymes inside cancer cells that control how they grow, divide and respond to their environment,” Flanagan says. “That gave us the idea that we could use those same cancer drugs to learn more about how brain cells control neuroinflammation.”

They tested this idea by screening compounds known to target cancer pathways in brain-cell models of neuroinflammation. This identified the involvement of a class of enzymes called cyclin-dependent kinases. Some of the enzymes in this family have long been studied in cancer research, with some emerging research showing effects in inflammation.

The team focused in on one of these enzymes, applying NeSI’s supercomputing power to complete a digital screen of a library of 118 million molecules against the enzyme’s 3D atomic structure.

The screen took two weeks to complete and delivered hundreds of candidates with the potential to turn off the enzyme in brain cells. Then, by testing fewer than 100 of these against patient-derived brain cells, the researchers identified one that was able to block an inflammatory response.

This validated the idea and created a foundation for the new Drug Discovery Programme to take shape. While this is a significant start, there are potentially thousands of other enzymes or proteins involved in neuroinflammation that we do not know about.

Conole, a chemical biologist at the University of Auckland's Department of Cancer Sciences, adds his expertise to illuminate what he calls the “dark matter” in brain cells.

“Brain cells contain at least 20,000 proteins, and while we understand parts of the complex cellular pathways that cause inflammation, we do not know all the proteins involved,” Conole says.

To track down the right targets in this ‘dark matter’, Conole uses special sticky molecules that block brain cell inflammatory responses, and a technique called precision mass spectrometry, to keep the research focused on the most important proteins.

Funded by both the Neurological Foundation and the Hugh Green Foundation, Conole and Dragunow, as well as PhD student Raahul Sharma and post-doctoral fellow Dr Caitlin Oyagawa, uncovered new targets for neuroinflammation using this chemical biology approach, which will be further pursued in this research programme.

“This work is important to making sure that we apply our digital screening approaches to the most relevant proteins,” Flanagan says. “Advances in AI technology to predict protein structure creates a really exciting opportunity to translate the outcomes of these chemical biology studies into the drug discovery pipeline.”

Crossing the barrier

Brain disease is notoriously difficult to treat, in large part due to the blood–brain barrier, which prevents viruses and infections from entering the brain, but also blocks most medicines. Even if the team find the one-in-a-billion molecule, it still needs to be delivered effectively to target brain tissue.

“There has been no lack of research efforts to find a drug to cross the blood-brain barrier and stop neuroinflammation, but therapies that show promise in animal models often fail to translate effectively to clinical settings,” Dragunow says.

When the in-vitro testing begins, a model of a bloodbrain barrier grown from human brain cells is ready to go. The remarkable model is the result of two years’ work by Dr Rebecca Johnson, a Research Fellow and Neurological Foundation First Fellowship recipient, and Dragunow Lab Manager Sheryl Feng. Johnson says, “The tissue samples are donated to the Hugh Green Biobank from people with brain diseases undergoing neurosurgery at Auckland Hospital.

“We’ve encountered a lot of complexity along the way. Humans are unique and every sample is variable and affected by factors like age, so we’ve put a lot of work into getting these models consistent.”

Research Fellow Dr Caitlin Oyagawa and Senior Research Technician Woo Lee in the Hugh Green Biobank will lead the testing that will help the team understand if, and how, the molecules they discover block inflammation and cross the blood-brain barrier.

Dragunow has been working with neurosurgeons since the 1990s to build the scientifically-valuable tissue collection. He says one of the Biobank’s key strengths is that it reflects New Zealand’s unique and diverse population. This ensures that local patients, reflecting Aotearoa’s rich heritage, are represented in the data, addressing potential differences in drug response not captured in predominantly European clinical trials.

“This was one of the two main reasons for establishing the Hugh Green Biobank,” says Dragunow. “The other is that, unfortunately, traditional approaches of testing drugs using animal models, or human brain cells that have been immortalised for research, have largely failed to deliver effective medications for brain diseases in people.

“We’ve spent about 15 years developing methods to generate advanced patient-derived brain cell models for drug testing that more closely reflect disease in people. We’ve also pioneered a screening method which accurately measures drug effects on those brain cells.

“Our approach is unique on a worldwide scale and driven by the generous brain tissue donors and their families as well as the wonderful neurosurgical teams at Auckland Hospital.

“We are now in a position, with this funding, to really make an impact and hopefully honour the wishes of the brain tissue donors by developing effective medications for brain disorders,”  Dragunow says.

Finding the one-in-a-billion

Leading the high-performance computing element of the project is Dr Jack Copping, a computational chemist trained by Flanagan who specialises in large-scale molecular screening. His work involves screening vast ‘compound libraries’ which can contain billions of molecules.

“There are libraries out there with tens of billions of molecules. While fantastically large, they represent only a small fraction of molecules that could be relevant for treating disease,” Copping says.

“We’re dealing with enormous datasets, and even a tiny error in a file can cause the whole system to fail, so a lot of our work at the moment is about making sure everything is clean and usable.”

Copping adds that NeSI is giving the research a globally unique leg-up. “We are fortunate that New Zealand has a national supercomputing system,” he says. “There are many places in the world that do not have that capability.”

As the scale of the search grows, so too does the need for smarter tools. Rather than simulating every molecule against the drug binding site on a protein, researchers are training neural networks (a type of AI model) to recognise patterns, learning which types of molecules are most likely to pass the digital screen.

“The idea is that we can use machine learning to predict which molecules are worth testing,” says Copping. “That allows us to expand the search upward from one billion and hopefully reduce the search time from months to days.”

The final stage – reaching patients

With clinicians already involved, including neurologists Dr Zoe Dyer and Dr Nicholas Child, and neurosurgeons including Mr Jason Correia and Mr Patrick Schweder, the pathway to patients is built in from the start. This ensures researchers consider the needs of clinicians treating brain disease.

Even more experts are poised to come on board, including drug development expert Dr John Villiger, who is an Honorary Associate Professor at the Centre for Brain Research and a board member on the University's commercialisation arm UniServices, is helping to map out the path to clinic.

Flanagan says,“It’s exhilarating to think how quickly we could move from a billion molecules to finding one that could be developed into a drug.” While the final answer may still be hidden in that vast chemical universe, the tools to find it are now firmly within reach.

This $1 million Drug Discovery Programme is funded by the Neurological Foundation as part of its Programmes and Platforms investment. The Foundation funding for research totals around $5m annually.