The use of artificial intelligence continues to grow in society, and its presence within Thoroughbred racing is also starting to increase.

Dr. Mary Robinson, assistant professor of pharmacology and director of the University of Pennsylvania’s Equine Pharmacy Laboratory at New Bolton Center, discussed how the Pennsylvania Equine Toxicology and Research Laboratory is using artificial intelligence to help improve the detection of drugs in post-race tests during PETRL’s year-end report at the Jan. 28 Pennsylvania State Horse Racing Commission meeting.

PETRL has developed methods to identify and quantify a large number of drugs in horse blood and urine and continues to develop novel methods for screening, confirming, and quantifying the presence of illegal drugs in racehorse plasma or urine samples.

Robinson shared how the use of AI has helped make the process easier to identify targeted drugs, for which PETRL’s confidential list contains hundreds.

“What (AI) does is look for key identifying characteristics for the drugs that are on the list to see if those are found on any of the samples that are put through the test,” Robinson said. “This alleviates our time spent manually going through all the data for each sample for every single drug. It’s been a huge improvement that we’ve been working on for several years.”

AI is also being implemented to detect untargeted drugs, those not registered on PETRL’s list, to identify anything the horse has been given, even though what was given may not exactly be known. This work has been spearheaded by Dr. Fuyu Guan, a research professor of equine forensic chemistry at UPenn’s School of Veterinary Medicine.

“(Guan) came up with a mathematical algorithm to be able to compare the control samples to horses that have substances in their sample and identify those substances without actually targeting and telling the machine to look for those systems,” Robinson said.

The algorithm takes two parameters that were identified to indicate peaks of interest.

“We need to hone in on which peaks are important, which are the ones that are unique that we want to know more about,” Robinson said. “We’re looking at the post-race samples that have things in them as well as the ones that don’t to try and help tell us what is normally found there and what is not.

“It’s kind of like taking the whole batch of data and then interrogating it to see what kinds of patterns the computer can find. At the end of the day, artificial intelligence is really about pattern recognition. The computer is identifying patterns in these large data sets and putting it all together for us in ways that our brains can’t easily do.”

The study spearheaded by Guan was published in a May 2021 edition of Analytical Chemistry.

In another project, Dr. Bethany Keen, whose postdoctoral research fellowship at UPenn is being supported by the Racing Medication and Testing Consortium and the Commission, has been using metabolomics to detect bisphosphonates.

Metabolomics is described by the National Institutes of Health as an emerging field and defined as the comprehensive measurement of all metabolites and low-molecular-weight molecules in a biological specimen. Bisphosphonates are a drug class that target osteoclasts, the cells responsible for breaking down damaged bones.

“By inhibiting the breakdown of bone, you can promote and increase in the amount of bone in that region,” Robinson said.

According to Robinson, bisphosphonates are approved for treating horses with navicular disease, a chronic degenerative condition that affects the navicular bone in the front hoof. Robinson described the bone as having holes that make it look like a sponge in horses with the disease. 

“If you use this drug, it can cause the bone to increase in the area,” Robinson said. “Unfortunately, we know that people have been using bisphosphonates in young horses as well—sometimes before the sales to try and make their X-rays look better. They are banned in racehorses and we have several tests we can use to look for them.”

However, detecting bisphosphonates can be difficult as they tend to stick in the bone and get released very slowly. This keeps the amount in the blood very low and hard to detect in post-race tests. Keen has developed a mathematical model that can help identify horses that have been treated with bisphosphonates from horses that have not.

“It’s not looking at the drug itself, but it’s looking at the effect of the drug on the body,” Robinson said. “It’s really exciting that we were able to see this model separate out horses that were treated from horses that weren’t treated. That gives us a potential of using this as another screening tool to look for horses that may have been exposed to bisphosphonates when we can’t actually detect the bisphosphonates itself.”