Creating the Only Stratification Test for
Triple Negative Breast Cancer
Stratification tests exist for HER2+ and ER+ Breast Cancer. However, a stratification test for Triple Negative Breast Cancer (TNBC) has remained elusive. Our journey to create a stratification test for TNBC consisted of three major steps. The first step was to understand the genes that could tell us how a patient with TNBC would respond to the standard-of-care treatment. Second, we applied machine learning to analyze how those genes would change in response to as many clinical variables as possible. Our last step was to ensure the test was reproducible in laboratories around the globe. Our journey through these steps produced over 17 scientific publications since 2001 each documenting a different stepping stone to creating the BA100 Triple Negative Breast Cancer Stratification Test.
Understanding Housekeeping Genes and
The first step of our journey was to identify the genes fundamental to normal breast tissue development that could predict which patients would achieve pathological complete response . This discovery wasn’t enough to create a stratification test, we also needed to identify which genes could predict the patients that would not achieve a complete response . After many years of understanding which genes were critical to stratifying the two groups of TNBC Patients we found a common thread – Housekeeping Genes .
The name says it all. Housekeeping Genes are responsible for the minute-by-minute clockwork of cells. Because Housekeeping Genes are ubiquitous, they are often overlooked as having a role in cellular abnormalities, like cancer. As it turns out, for Triple Negative Cancer Patients, Housekeeping Genes are indicators of treatment response. When Bioarray research indicated that Housekeeping genes might be a unique indicator of a patient’s treatment response the team explored the family of Housekeeping Genes even further. The final results say it all. Of the 325 gene panel created by Bioarray, 77% of them are not found in any of the nine common cancer panels. Of those unique genes, 102 of them are Housekeeping genes . What we learned was that, in Triple Negative Breast Cancer, if these Housekeeping Genes are downregulated (shut down) by a patient’s cancer, it is unlikely that they will achieve a complete response to the standard-of-care chemotherapy.
Applying Machine Learning
Although the group of genes that could identify which TNBC patients would and would not respond to treatment was critical, we needed to further refine the group based on all of the clinical variables that a TNBC patient can go through. Each TNBC case is unique and the utility of our test needed to persist regardless of these variables. To help us understand the almost limitless impact of these variables we used machine learning.
Machine learning requires two things. The first is an algorithm, or program, that the computer uses to translate information into an actionable result. The second is a continuous flow of information to feed the algorithm. The success of machine learning depends on the sophistication of the algorithm and the quality of information. For Bioarray Genetics this meant having some of the best bioinformaticians in the world write an algorithm that would be able to take mass amounts of complex clinical and genetic data and identify the genes critical to the desired patient outcome. This also meant we needed a continuous flow of critical clinical data (e.g. patient demographics, treatments, cancer size, gene profile).
Ensuring Test Reproducibility
The last phase of creating the TNBC stratification test was to ensure that it was reproducible regardless of when, where or who ran the test. Identifying the right gene sets isn’t enough if the test couldn’t be run around the world. We created a study to test the reproducibility and what we found was that our gene assay was highly reproducible between multiple days and users . This reproducibility was the final step in being able to bring the BA100 Triple Negative Breast Cancer Stratification Test to market.
The journey to discovering, developing and commercializing the BA100 Triple Negative Breast Cancer Stratification Test took 8 years. However, we now have a collection of 325 unique genes and an incredibly powerful machine learning algorithm that can continue to predict new applications for those genes. Taken further, our algorithm can continue to consume new clinical data to help us develop future tests based on changing treatment options. We have the data and genes to keep making an impact on areas of unmet medical need in breast and other cancer types.