Jefferson Investigates: AI Tool to Differentiate Brain Tumors, New Clues to ALS, RNA subtypes and Prostate Cancer

When an MRI shows a mass in the brain’s parasellar region, physicians must determine the tumor type to recommend treatment. A study from Jefferson Health researchers shows that automated machine learning (AutoML), a user-friendly AI-powered tool, can recognize subtle differences between benign tumors.

Pituitary macroadenomas (PA) and parasellar meningiomas (PSM) are the most common benign tumors in the parasellar region, near the pituitary gland. They can look very similar in MRI images, but there are different surgical procedures for each. Presurgical biopsy is uncommon, so identification is crucial.

“The more certainty we have about a tumor before going in, the better we can counsel patients and prepare for surgery,” says otolaryngologist Gurston Nyquist, MD, the study’s senior author.

AI-powered machine learning and deep learning have used data-entry images to diagnose PAs and PSMs with high accuracy, but they’re too complex to be used widely. (Most clinicians can’t write software enabling AI to interpret images.) AutoML requires little except data entry.

“It does learning on its own, as opposed to you telling it what to do,” Dr. Nyquist says. “It’s looking for subtleties that we don’t pick up on.”

Previous research has used AutoML to differentiate between brain tumors, but this is the first AutoML model specifically trained on parasellar region masses.

“We wanted to look at tumors that are almost in the exact same space that look very similar,” Dr. Nyquist says.

Researchers prepped AutoML with tumor images from a publicly available, multi-validated dataset. AutoML then identified PAs and PSMs in 116 Jefferson patients with precision greater than 99%. The researchers are now planning a multi-institutional trial.

“We want to see if we can consistently get high confidence intervals with an increased number of patients and data, then potentially open it up to clinical use,” Dr. Nyquist says.

Sidney Kimmel Medical College students Elliott Sina, Emma Anisman, Nickolas Pudik and Emma Tam; research fellow Kelsey Limage; resident Chase Kahn, MD; otolaryngology clinical fellow Srihari Daggumati, MD; otolaryngologists Mindy Rabinowitz, MD, and Marc Rosen, MD; and neurosurgeon James Evans, MD, contributed to the research.

By Lisa Fields

Amyotrophic lateral sclerosis (ALS) is a devastating neurodegenerative disease with no cure and limited treatment options. One of the earliest signs of ALS is overactive brain signals known as cortical hyperexcitability. This activity appears even before motor neurons begin to degenerate and physical symptoms such as trouble walking or swallowing show up. Now, Thomas Jefferson University researchers have discovered that neurons carrying the most common genetic cause of ALS respond abnormally to cortical hyperexcitability. The findings offer fresh insight into how the disease takes hold.

The team focused on the most common genetic cause of ALS, a nucleotide repeat expansion in a gene called C9orf72, colloquially dubbed C9. To figure out how C9 affects brain cells known as cortical neurons, the researchers first turned stem cells from ALS patients into cortical neurons in a dish. Then, they asked how these cells react when stimulated in a way to mimic cortical hyperexcitability.

“The question was, do cells with the C9 mutation respond differently to the same stimulation? And the answer is yes, they do,” says Layla Ghaffari-Starr, lead researcher of the project and now Director of Disease Modeling at Synapticure, a health care provider for neurodegenerative diseases like ALS.

The scientists, members of Sidney Kimmel Medical College, found that even at baseline, C9 neurons were genetically different. When the researchers stimulated the cells, the C9 neurons activated divergent genetic programs compared with healthy ones. Some typical pathways were maintained, but others failed to turn on. In addition, aberrant programs were triggered, including some previously linked to ALS.

The findings may help explain the trajectory of the disease from early hyperexcitable states to later neuron loss, and could point to biomarkers for detecting ALS sooner.

“Our research serves as a resource to the ALS and scientific communities to look at dysregulated genes in C9 mutant neurons,” says Dr. Ghaffari-Starr. She and neuroscientist Aaron Haeusler, PhD,  have created an interactive dataset for researchers and the ALS community to identify therapeutic targets.

“It really takes everyone to solve this,” says Dr. Haeusler, who is especially grateful to the patients who donated their cells for research. “Without them, we wouldn't be able to make these advancements.”

By Roni Dengler

Prostate cancer is the second-most common cancer in men. A new study from Thomas Jefferson University uncovered a new potential therapeutic target in tiny molecules called tRNA halves.

Transfer RNA (tRNA) are molecules that transport specific protein-building blocks called amino acids to help build full proteins. tRNA halves are small fragments formed when tRNAs are cut in half, often in response to cellular stress or sex hormones, like androgens, which drive prostate cancer. A study from 2015 reported that tRNA halves are highly expressed in prostate cancer.

“Because tRNA halves are so abundant in prostate cancer, we thought they may be doing something important,” says biochemistry researcher and senior author of the study, Yohei Kirino, PhD.

In their new study, Dr. Kirino and his team discovered that tRNA halves can help prostate cancer cells grow. The researchers focused on one particular tRNA half that was especially abundant in the prostate cancer cells they analyzed. They found that this tRNA half acts almost like a switch to turn on more cell growth. The tRNA half can do this by reducing the amount of a protein called p21, which normally serves as a brake to slow down cell growth. When p21 levels drop, cells are free to grow faster and the cancer progresses.

Digging deeper, the team studied how the tRNA half was able to weaken this cellular brake. Under normal circumstances, a protein called YBX1 allows p21 to do its job to repress cell growth. However, when the tRNA half is present, it interferes with YBX1 and as a result, p21 levels go down, releasing the break on cell growth and allowing cancer cells to grow faster.

“tRNA halves are some of the most abundant molecules in prostate cancer cells, yet they are severely understudied,” says Dr. Kirino, a member of Sidney Kimmel Medical College. “Our work shows that these RNA fragments are not simply byproducts of RNA breakdown, but powerful regulators that can influence cancer growth. Understanding their impact could open up new possibilities for therapies in hormone-dependent cancers.”

By Moriah Adde