UCLA’s Hosein Mohimani matches molecule and machine learning to better understand metabolism

Hosein Mohimani describes two of his signature research projects as coming from "two different universes" within the biosciences. One seeks new weapons against treatment-resistant infections, and the other explores the role of a specific family of enzymes in drug processing.

The common factor that links these investigations is actually the tools used to pursue them. Mohimani, an associate professor of computational medicine in the David Geffen School of Medicine at UCLA, connects machine learning - a type of artificial intelligence in which algorithms identify patterns after being "trained" using plentiful data - with mass spectrometry - an analytic method for revealing the size and identity of molecules in a sample. The combination has the potential to be a sort of Swiss Army knife for addressing many of the mysteries of biology.

In an interview, Mohimani, a member of the California NanoSystems Institute at UCLA who came to Westwood in 2025 from Carnegie Mellon University, discussed the two universes in his research and where he sees it all going.

How do mass spectrometry and machine learning enable your own work in drug discovery?

There are many different types of bacteria, fungi and plants that evolved to produce molecules that can kill the most dangerous pathogens. But the molecules are in complex mixtures, and nobody knows what they are. It's difficult to detect them.

We use mass spectrometry to screen these natural product extracts. Our second signal is genomic data about the organisms that produced them. With those inputs, we use machine learning to find out the structure of the molecule, whether it's been identified before, and whether it has antibiotic activity or newly mapped substructures that haven't previously been tested for activity against pathogens. That paves the path toward isolating those molecules and testing them.

At Carnegie Mellon, we screened a few thousand bacterial strains and identified hundreds of previously unknown natural products, then tested them against the deadliest human pathogens. That led to the discovery of a molecule with a previously unreported mechanism of action that kills Candida auris, a drug-resistant, fast-spreading fungal pathogen detected in hospitals across the United States.

I founded Chemia Biosciences Inc. with a project scientist who was leading this work, and the company is developing a drug based on that molecule. Chemia has also established a collaboration trying the methods we developed on a much wider range of molecules and seeing if their compounds can kill fungal disease in crops.

How did you get into studying drug metabolism?

Antibiotic discovery was my sole focus for about 17 years. Recently, I've moved toward answering other interesting questions in life sciences. When I was first coming to UCLA, I spoke with Thomas Vallim [associate professor of biological chemistry and CNSI member] about cool things that he's doing in his lab.

He works to understand how the body metabolizes fats and other lipids using a group of enzymes called cytochromes P450. How P450s change lipids in our body has important implications for our health, and a mutation in one of these enzymes could lead to disease.

A related question is how the drugs we consume are changed by P450 enzymes. Variation in P450 genes across the population causes differences in how drugs are metabolized, leading the drugs to lose potency in some people and produce toxic byproducts in others.

In drug development, the Food and Drug Administration asks about the products of P450 metabolism that have a yield of 20% or higher. But our studies have shown there are a lot of other products we don't know about that are usually lower yield, but could be really toxic or really potent.

In the past, the limitations of lab technology made it impossible to analyze these products. But in the past nine months, we've screened 2,000 drugs for all the bioactivity of one P450 enzyme using high-throughput mass spectrometry and machine learning. In the next few months, we want to go after all the P450 enzymes that are present in humans.

Usually, a single molecule is tested at one time. Studying P450 metabolism, we're instead analyzing tiny amounts of thousands of compounds. There are ways to remove complexities such as interactions, and this method can make screening hundreds of times cheaper.

Where do you see this research going?

We want to figure out how long it takes for a drug to be cleared from a patient's system, which other molecules the drug could transform into and how easily it gets through cell membranes. All these questions can be answered using mass spectrometry.

We're also interested in training a machine learning model that can look at an enzyme sequence and the structure of the molecule it's modifying, and then predict the likely products. As we understand more about how enzymes change molecules, we could train a machine learning model to plan how they can be used to get desired chemical products in drug development and beyond.

Why is UCLA a good place to do this work?

The quality of my collaborators at the Geffen School of Medicine is top-notch, and I have access to great mass spectrometry instruments. Robert Damoiseaux [professor of molecular and medical pharmacology and of bioengineering], who runs the Molecular Screening Shared Resource at CNSI, has been a crucial collaborator. The automation there really speeds up the work.

I would love to establish more collaborations. I think there are a lot of cool projects where we can apply our knowledge of mass spectrometry and machine learning.