A new tool from the lab of Velia Therapeutics co-founder Alan Saghatelian, Ph.D., has proven capable of scouring genetic data for tiny proteins that may serve as new therapeutic targets, opening the door for more exploration of the role these microproteins can play in medicine.
When given old gene expression data sets taken from healthy and cancerous lung tissue, the model found several potential microproteins—typically defined as proteins that are fewer than 100 or 150 amino acids long—that were much more common in tumors. Comparing these data to existing protein profiles of tumors confirmed that the petite proteins are, in fact, made by cancer cells.
“If any of those microproteins turn out to be cell-surface expressed, you could develop antibodies or other ways to target those,” Saghatelian, a professor at the Salk Institute for Biological Studies in San Diego, told Fierce Biotech. “Or, if they're intracellular and they're maybe driving the tumor, you can figure out how to drug that pathway.”
This case study highlights the potential utility of ShortStop, the model spearheaded by diehard Chicago Cubs fan Brendan Miller, Ph.D., a postdoctoral researcher working with Saghatelian. Details of the model were published in BMC Methods on Aug. 1.
Since the full human genome was first published in 2001, scientists have known that much of our DNA is so-called “dark matter” of unclear function, including some sections that seem like they provide instructions for making extremely small proteins. At the time of the Human Genome Project, scientists assumed these regions just happened to look like small genes but didn’t actually make any proteins.
About a decade later, a group led by Jonathan Weissman, Ph.D., then at the University of California, San Francisco, debuted a technique to profile ribosomes, the protein-making factories of cells, to see all the proteins they produce. This revealed that cells do in fact turn those mysterious DNA segments into many different microproteins. Saghatelian’s group then showed that you could detect these proteins using a technique called mass spectrometry.
Since then, researchers have been toiling to find more microproteins and understand the roles they play in cell biology. When Miller joined Saghatelian’s group, he realized that they now had enough data on what real microprotein genes look like to train a computer model to recognize them.
But, for the model to be able to accurately identify microproteins, it also needed to be trained on false positives—sections of DNA or RNA that look like they could correspond to a microprotein but don’t actually make anything. And, while the team had plenty of data for genuine microproteins, no one had ever bothered to collect all the potential microproteins that didn’t turn out to be real.
To get around this, Miller generated 1,000 data sets of random genetic sequences that match the length of microprotein genes. When presented with real data, the model could then judge whether a putative microprotein looks more like a random jumble of DNA or the real thing.
“We think the approach is terrific,” said John Lepore, M.D., president and CEO of ProFound Therapeutics. Given the difficulty of sorting through which potential microprotein genes actually make proteins, he said, “the more we have tools and methods filtering down and figuring out which ones are going to be more important, the better for the field and the better for patients.”
Saghatelian’s team is now putting ShortStop to bat in a variety of disease areas. Miller is using it to probe potential Alzheimer’s disease targets, and fellow Salk professor Satchidananda Panda, Ph.D., is collaborating with Saghatelian to look for microproteins in muscle.
“The beauty of ShortStop is that it's pretty much sample agnostic,” Saghatelian said. “If you can get RNA-seq data sets, which there are tons out there, you can go after anything you want.”
But before he’d consider starting another company, Saghatelian said he’d need to see a clearly identified target to pursue. His prior microprotein venture, Velia Therapeutics, announced it was winding down in January, and Saghatelian said the science was just too early.
“We started the company without really bringing in any patents or technologies from the lab,” he explained. “There wasn't the one target that they were going to start a program with right away.” In the current market, Saghatelian said, companies don’t have enough runway to sustain a basic science program without any identified development prospects.
Whether a company’s approach is worth the wait ultimately comes down to its specific investors, Lepore said. His company, Flagship Pioneering-backed ProFound, debuted in 2022 and is also exploring the hidden proteins of the human genome’s “dark matter.”