Meenhard Herlyn, D.V.M., D.Sc., and Noam Auslander, Ph.D., are leveraging cutting-edge computational tools to better understand the genetic drivers of melanoma

Melanoma, the deadliest form of skin cancer, can be challenging to treat because much remains unknown about what causes cancer cells to grow and spread and many melanoma tumors are resistant to treatments.

The Wistar Institute’s Meenhard Herlyn, D.V.M., D.Sc., and Noam Auslander, Ph.D., have forged an innovative, multidisciplinary partnership to tackle this problem through their combined expertise in melanoma biology and machine learning. Their research improves our understanding of the genetic drivers of melanoma – and provides a vital foundation for identifying new therapeutic targets.

While recent advancements in treatments have transformed the treatment landscape and provided cures for some people living with advanced melanoma, approximately 50% of melanoma patients still don’t respond to the current options.

Researchers have struggled to develop more effective treatments because melanoma is one of the most complex and highly mutated cancers. With such variability in how tumors mutate – it’s difficult for scientists to pinpoint which biological factors should be targeted through different treatments.

Through combining their expertise in computer science and melanoma biology, Herlyn and Auslander are leveraging cutting-edge computational techniques, or bioinformatics, to identify promising new biological pathways to explore for future treatments.

“We’re using bioinformatics to try to identify the pathways and biological candidates that hold major meaning for the fate of a tumor – and the most effective treatments for it,” Herlyn explained.

Herlyn has worked in the field for 50 years and is all-too familiar with the challenges of melanoma. His lab has earned a reputation as one of the best-known research groups in the world through their extensive work on melanoma biology. But no matter how many papers he publishes, or how many decades he works in the field, he knows he’s limited as a biologist in his ability to probe the high variability of melanoma tumors.

“With melanoma having so many mutations, we don’t know which are important and which are not,” Herlyn explained. “This can’t be solved by biologists alone because it’s way too complicated to study at a single-cell level, which is the traditional method to examine the genetic drivers of cancer.”

Herlyn explains that in order to test whether a certain mutation is activating a gene, a biologist would have to investigate it on a singular cell level by taking the gene, mimicking the mutation, testing it against normal cells, and seeing whether it drives cancer development. But with each individual tumor carrying the possibility of hundreds of different mutations; and thousands of different data points to consider across DNA, RNA, and protein markers; it’s a daunting problem to determine which of these mutations is a potential driver of cancer. Especially when testing a single gene can take an entire year, or more, to research.

“That’s why I turned to a bioinformatician to find patterns of reactivity,” he said.

Auslander is a computer scientist who specializes in developing machine learning methods to better understand the factors driving cancer development – and the patterns that can inform more effective diagnosis and treatment. She can take massive amounts of data, like the genetic signatures of hundreds of tumors with hundreds of potential mutations and use advanced computational tools to detect patterns in the data that could never be uncovered through any one scientist.

“Melanoma is an ideal cancer for bioinformatics because it has one of the highest mutation burdens of any cancer,” said Auslander. “That wealth of mutations provides a tremendous opportunity to uncover patterns that would be impossible to detect manually.”

Herlyn’s lab had the high volume of data Auslander would need for her analysis thanks to a decade-long effort to collect tumor samples for research. Through partnerships with eight institutions including the University of Pennsylvania, UT MD Anderson Cancer Center, and Massachusetts General Hospital, his lab collected and analyzed over 600 tumor samples from their collaborators.

Because the data had been collected over many years, and across many different medical centers, Auslander’s first major technical challenge was to process, harmonize, and integrate data from 242 tumor samples.

She then used advanced computational tools, including Artificial Intelligence (AI), to model how different mutations shape molecular behavior in a tumor environment. She integrated and analyzed thousands of data points across genomic, molecular, and structural, and clinical data and conducted side-by-side comparisons across several different computational approaches to look for patterns.

Through the rigorous, computational analysis, the pair were able to identify promising patterns and uncover new aspects of melanoma biology. Their findings offer critical clues into how and why certain mutations cause cancerous cells to grow – and resist treatments. Biologists, like Herlyn, can now prioritize studying the specific genes that could be connected to a tumor’s progression.

“Through this approach, we were able to systematically link specific mutations to downstream biological effects, providing a valuable framework for identifying clinically relevant biomarkers and therapeutic opportunities,” Auslander said. “This project highlighted how computational methods can reveal new biology that would otherwise remain hidden.”

Auslander and Herlyn are sharing all of their findings in an open-source format to inform future research projects – and help advance the field of melanoma biology.

“What does melanoma research need? It needs new targets, new therapies, and totally new approaches,” said Herlyn. “I think our only hope of progress is through multidisciplinary research, because one laboratory or one investigator cannot singlehandedly do the work that’s needed to advance melanoma research.”