Machine Learning Tools Detect New Brain Cancer Cell Types

Researchers from Vanderbilt University leveraged unsupervised and automated machine learning techniques to analyze millions of cancer cells and identify new cancer cell types in brain tumors.

The team’s findings hold important implications for treatment of glioblastoma, an aggressive brain tumor with high mortality, as well as the application of machine learning to broader cancer research.

In a study published in eLife, researchers described the development of the Risk Assessment Population IDentification (RAPID) tool, an open-source machine learning algorithm. The tool revealed coordinated patterns of protein expression and modification associated with survival outcomes.

To develop the tool, researchers used data on cellular proteins that govern the identity and function of neural stem cells and other brain cells. The data type used is called single-mass cytometry, a measurement technique typically applied to blood cancer.

After RAPID completed its statistical analysis and had found the needles in the haystack, only those cells were studied.

“Without any human oversight, RAPID combed through two million tumor cells – with at least 4,710 glioblastoma cells from each patient – from 28 glioblastomas, flagging the most unusual cells and patterns for us to look into,” said Rebecca Ihrie, associate professor of cell and developmental biology at Vanderbilt.

“We’re able to find the needles in the haystack without searching the entire haystack. This technology lets us devote our attention to better understanding the most dangerous cancer cells and to get closer to ultimately curing brain cancer.”

The machine learning analysis allowed the team to study multiple characteristics of the proteins in brain tumor cells in relation to other characteristics, resulting in new and unexpected patterns. The tool was also able to discover brain tumor cells without added context, the researchers noted.

“One of the most exciting results of our research is that unsupervised machine learning found the worst offender cells without needing the researchers to give it clinical or biological knowledge as context,” said Jonathan Irish, associate professor of cell and developmental biology and scientific director of Vanderbilt’s Cancer & Immunology Core.

“The findings of this study currently represent the biggest biology advance from my lab at Vanderbilt.”

This study has broader implications beyond cancer research, the team stated. The data analytics techniques could be applied to wider human disease research and laboratory modeling of diseases using multiple samples. The study also demonstrates that these complex patterns, once found, can be used to develop simpler classifications that can be applied to hundreds of samples.

Researchers studying glioblastoma brain tumors will be able to refer to these findings as they test to see if their own samples are comparable to the cell and protein expression patterns discovered by the Vanderbilt team.

“The collaboration between our two labs, the support that we received for this high-risk work from Vanderbilt and the Vanderbilt-Ingram Cancer Center (VICC) and the fruitful collaboration with neurosurgeons and pathologists who provided a unique opportunity to study human cells right out of the brain allowed us to achieve this milestone,” said Ihrie and Irish.

In the journey to better understand and treat cancer, researchers have developed and leveraged machine learning tools to analyze more data. In October 2019, a team used a machine learning algorithm to detect malignant thyroid nodules in ultrasounds with 77 percent accuracy, providing a fast and inexpensive way to screen for thyroid cancer.

“Currently, ultrasounds can tell us if a nodule looks suspicious, and then the decision is made whether to do a needle biopsy or not,” said Elizabeth Cottril, MD, an otolaryngologist at Thomas Jefferson University, and clinical leader of the study.

“But fine-needle biopsies only act as a peephole, they don't tell us the whole picture. As a result, some biopsies return inconclusive results for whether or not the nodule may be malignant, or cancerous, in other words.”