One Step Closer to Developing a Cure for Melanoma
One Step Closer to Developing a Cure for Melanoma
May 27, 2020
The Jerusalem Post — In a scientific breakthrough, a research team, headed by Dr. Assaf Zaritsky from BGU’s Department of Software and Information Systems Engineering, has identified characteristics of melanoma cells that are likely to metastasize to other parts of the body. The research is a first and important step in developing novel treatments that could lead to a cure for melanoma–an aggressive form of skin cancer.
Dr. Zaritsky began the research during his postdoctoral studies at the University of Texas Southwestern Medical Center at Dallas with Gaudenz Danuser. The first stage of their research allowed doctors to predict what is likely to happen to the melanoma cancer cells and treat them accordingly.
This newly revealed second stage of research helps identify the properties of these cells that metastasize, “so we can create new treatments and eventually cures… If we know the properties of the cell that is going to metastasize, we can look for drugs,” says Dr. Zaritsky.
Normally, metastatic progression is predicted through a combination of genetic tests and patient history and static histological slides, which would not give information about the changes happening within the cells.
“The dream is that a person would come with stage III melanoma and doctors could predict if it would progress to stage IV or not and, based on that, adjust his or her treatment,” says Dr. Zaritsky.
Using deep neural networks – sophisticated mathematical modeling to process data in complex ways – Dr. Zaritsky’s team created a representation of the functional state of individual cells that can help predict the chances that a stage III melanoma will progress to stage IV, the most advanced phase of melanoma and a serious form of skin cancer, indicating that cancer has spread from the lymph nodes to other remote organs.
By computationally generating cell images that have never before been observed experimentally and by exploiting temporal information from live cell imaging experiments, the team reverse engineered the physical properties of the hidden image information that discriminates melanoma cells with low versus high metastatic efficiency.
The findings revealed that those cells that are likely to metastasize have pseudopodial extensions or miniature protrusions as well as increased light scattering.
“Deep neural network machine learning is a very powerful tool and can identify hidden patterns in complex cell imaging data that we do not see with our eyes,” Dr. Zaritsky explains.
However, he says that these techniques are often criticized as uninterpretable black boxes, lacking the ability to provide meaningful explanations for the cell properties that drive the machine’s prediction.
Using melanoma cells from patients that were previously implanted into mice, Dr. Zaritsky’s team demonstrated associated metastatic potential to the patient’s outcome. The team investigated whether this potential can be predicted from patterns in the cells’ appearance.
Dr. Zaritsky emphasizes that his work is “a long way from a cure, but now we can start thinking about it much more than we could before.”
Additional researchers include Andrew Jamieson, Erik Welf and Andres Nevarez in Texas. The research was published on bioRxiv.org and has been submitted for evaluation in a peer review journal.