College of Pharmacy develops computational core at WesternU

Advances in the speed and power of computers and artificial intelligence are opening new frontiers in the understanding and cataloging of biological and pharmaceutical information. As high-powered programs process billions of data points, novel and groundbreaking drug treatments are waiting to be discovered.

Three Western University of Health Sciences College of Pharmacy professors are at the forefront of a computational core – utilizing computers to advance drug discovery, protein dynamics, bioinformatics and more.

COP’s introduction to computational core came in 2013 with the hiring of Associate Professor Yun Lyna Luo, PhD, who completed her doctoral and postdoctoral training in computational chemistry and biophysics. She established WesternU’s first high-performance research computing lab to investigate the structural dynamics of complex biomolecules and conduct mechanism-based drug design. Among her ongoing NIH-funded projects are gating mechanism study and drug discovery of mechanosensitive Piezo channels, gap junctional channels, and CatSper channels. Together with her current research team (Dr. Aashish Bhatt, Dr. Deepak Kumar, Yichun Lin, Chenyun Wen, and Mekedlawit Efrem), they work closely with experimental collaborators and develop computational methods and software to accelerate biomedical research. 

“I have been here 11 years, and now it’s overwhelming the number of people who want to collaborate with us to use our computational skills,” Luo said. “We are using the computer and AI to understand how protein functions and how to find new drug candidates.”

The field of computational-aided design has exploded in the past five years for two reasons: computing power is increasing exponentially while costs are reducing, and artificial intelligence starts to become an indispensable tool in basic research, Luo said.

“Five years ago, we knew a limited number of protein structures,” Luo said. “Now, AI can predict protein structure, so we have an almost infinite number of protein structures we can simulate.”

The computing power of AI can accelerate drug design and reduce costs, Luo said. A computer can identify potential drugs early in the drug discovery pipeline without having to synthesize molecules.

“That’s why in recent years, capacity has improved so much. We have a real impact in the field,” Luo said. “We recently published a paper in PNAS in which we used computer modeling to find two new molecules that activate Piezo1 channel.”

Two of Luo’s previous post docs (Dr. Wesley Botello-Smith and Dr. Wenjuan Jiang) were recruited to pharmaceutical start-up companies to develop algorithms to predict protein structure and protein drug binding. The training Luo provides to her postdocs and students is helping them be more competitive in the job market. With so much venture capital money invested, real products are coming out with reasonable speed and lab performance.

“I’m expecting to see in five years there will be a drug in clinical trial that was purely developed from a computational lab,” Luo said.

Dr. Luo’s work received continuous NIH Research Project Grant (RO1) funding, which resulted in building COP’s computational core by acquiring a large computer cluster, said COP Biotechnology and Pharmaceutical Sciences Department Chair Stephen O’Barr, PhD. She has gone on to collaborate with multiple institutions (Yale University, University of Chicago, UC Davis, City of Hope) and companies who request her expertise through sub-awards and contracts.

COP Associate Professor Bradley Andresen, PhD, FAHA, is using machine learning to find new compounds to target orphan GPCRs (G protein-coupled receptors), which are the largest single target in the pharmacopeia. These heptahelical receptors are essential to many fundamental physiological processes, such as vision, taste, and smell, as well as nerve transmission, the heartbeat, and GI functions, to name a few.

The human genome displayed about 800 GPCRs; however, nearly 100 of these GPCRs are considered orphans. An orphan GPCR is a GPCR with no known physiological ligand. We know from genetic studies (screens of genes and knockout animal studies) that orphan GPCR expression levels are altered in disease, and the knockout animal studies provide hints that these GPRCs may contribute to disease, Andresen said.

However, studying a GPCR is like studying a lock – you need the key or a lock pick, which for GPCRs is ligands.

“My laboratory uses cutting-edge tech to screen 1.7 billion drug-like compounds to find novel ligands for orphan GPCRs,” Andresen said. “Those ligands will then be examined using pharmacological stone age tools to determine if the ligand interacts with the receptor and if it can turn it on (an agonist) or off (an antagonist). The early tools are applied to this project because when those tools were developed, we had no method of visualizing the ligands, allowing functional screens to be conducted with small amounts of a novel compound.”

To study these different receptors, you need to be able to turn them on and off.

“Without turning them on and off, you can’t say how they play a role in what’s going on,” Andresen said. “I see myself as a blacksmith, making tools to figure out what receptors do. Could these become clinical drugs? That’s not the goal. It’s a lofty goal, but first I just want to make the tools to understand them. Physiologists and pharmacologists can then use those tools to figure out what those tools do in disease states.”

COP Associate Professor Peter Oelschlaeger, PhD, MS, received an intramural grant to develop a website for metallo-beta-lactamase amino acid renumbering. Enzymes able to hydrolyze and cause resistance to beta-lactam antibiotics are called beta-lactamases and are grouped into four classes (A-D). Class B beta-lactamases (BBLs) employ zinc ions and are also known as metallo-beta-lactamases (MBLs). BBLs have a unique conserved structure including zinc ligands, which has prompted researchers to develop the BBL standard numbering scheme. Because segments between conserved residues can vary, the standard numbering can be quite complex and require insertions and deletions. As a result, renumbering all amino acids in the enzyme sequence manually is tedious, and recently the BBL numbering has been utilized less, especially since the advent of the New Delhi MBL (NDM) enzymes.

“Therefore, we have written Python programs that can automatically renumber NDM amino acids in sequence (FASTA) and structure (PDB) files to conform with the BBL standard numbering,” Oelschlaeger said. “The goal of this grant is to expand these programs to other BBL enzymes and make them freely accessible by designing a website.”

Beta-lactams such as penicillin constitute more than half of all antibiotics. They owe their antibacterial effect to their ability to inhibit the bacterial enzyme transpeptidase, which is essential for the biosynthesis as well as the remodeling and repair of peptidoglycan, the major component of the bacterial cell wall. In addition, they are characterized by low toxicity because humans do not have a functional analog of this enzyme that could cause off-target effects, and most beta-lactams are not metabolized and excreted renally.

Each enzyme has a long chain of amino acids linked to each other. There are challenges to naming beta lactamases and numbering their amino acid residues. When a new beta-lactamase is discovered, the amino acid residues are numbered from 1 to however long the chain is.

Researchers are trying to design inhibitors that can then inhibit the enzyme so the antibiotic will still be effective. In order to show how a new molecule works, the researchers would use many different metallo-beta-lactamases to see if they work with different enzymes. Often they will make a basic crystalized 3D structure of protein to look at what residues interact with the inhibitor. Because of the numbering system, it’s hard to figure out a common way of how an inhibitor binds to the enzymes. That’s why this standard numbering of these amino acids would be beneficial.

“It would be too tedious to do it manually. That’s where the computer program comes in,” Oelschlaeger said.

He has a computer program that works for one of the families of metallo-beta-lactamase.

“My program only works for one. It searches for specific sequences, all the enzymes a family has, like a fingerprint. If you go to another family, it’s not going to work. Eventually the idea is to make it more generally applicable to different families — new enzymes from new families that haven’t been characterized yet,” Oelschlaeger said. “It’s just on my computer. Other people can’t use it. The idea of this intramural grant is to design a website where people could submit a sequence of a new metallo-beta-lactamase and the website would return the renumbered sequence.”

Eventually, everybody who comes out with new inhibitors will want to publish them. They will want to show figures how the inhibitor binds to the enzyme, Oelschlaeger said. It would be much easier if all enzymes have the same amino acid numbers for presentations.

For the bigger picture, this is just one enzyme class that happens to be important to antibiotic resistance.

“Almost all diseases have to do with some enzyme or protein. All the proteins have different amino acid numbering,” Oelschlaeger said. “If we standardize them – the same residues, interacting with ligands – it would be easier to compare similar enzymes. If we can show ‘you can do this,’ then we can transfer that knowledge to different proteins important for other diseases.”

“It’s exciting to watch how our department faculty leverage computational resources, backed by strong federal and industry partnerships, to accelerate drug discovery and development,” O’Barr said. “To amplify this impact, evolving our core into a University Center will further expedite discovery and development, propelling the University’s research reputation to new heights.”