Approach

Our vision is to improve care of cancer patients by developing biomarkers that combine the genotype and phenotype of tumors in treatment decisions, leading to a better ability to predict the course and therapeutic sensitivities of the disease. It is known that cancer evolves during treatment to acquire resistant traits. Particularly among patients with later-stage and metastatic disease, treatments often fail to account for the tumor’s adaptations and are inadequate, as pre-treatment biopsies are commonly used for therapy decisions upon subsequent relapses. It is time to match a tumor’s complexity with mathematical approaches and “big data” that are together capable of unraveling the diversity of cancer traits key to treatment response.

We have developed a biomarker that measures a tumor’s response to common breast cancer treatments using machine learning and genomic data. Breast cancer is the most common cancer worldwide and the second leading cause of cancer death in American women. Approximately 13% of women born in the US today will develop breast cancer at some time in their lives. Further, there are approximately 290,000 people diagnosed with invasive breast cancer each year in the US, and 80% are estrogen/hormone receptor positive. Therapies targeting estrogen signaling are commonly prescribed during years of treatment. Most progressive breast cancer tumors that are estrogen receptor positive ultimately develop resistance to endocrine therapy and become estrogen independent; however, no biomarker exists to match patients to endocrine therapy in progressive breast cancer.

To address this clinical need, we developed a biomarker for endocrine therapy response called “ENDORSE”. This model includes genes associated with long term survival after endocrine therapy as well as a pathway-based signature of genes expressed in response to estrogen. This biomarker discriminates women who respond to or are resistant to therapy with high significance. Five additional patient cohorts were assessed, and the ENDORSE biomarker demonstrated superior and consistent performance of the model over clinical covariates and multiple published signatures.

To implement the ENDORSE biomarker, we have opened a clinical trial, which focuses on second line treatment of metastatic estrogen receptor positive breast cancer patients. For these patients, there are multiple approved drugs; however, the only available test is for a PI3K mutation, which matches patients to PI3K targeting therapy, alpelisib. In SPOCK, the ENDORSE biomarker, as well as a second biomarker we developed for mTOR inhibitor response, will be tested for improved patient progression free survival outcomes.

Unravel Genomics has developed a fully automated, cloud-based, end-to-end biocomputational pipeline for the processing and analysis of patient genomic data. The different stages in the automated pipeline include collection and storage of raw sequencing data, preprocessing and quality control, machine-learning driven (AI-enabled) predictive analysis and delivering recommendations and analytics in a detailed graphical report.

Our biomarker workflow automates and streamlines the processing of raw DNA- and RNA- sequencing data. The RNA-seq processing is performed based on current best practices using the STAR pipeline and produces normalized counts from raw sequencing reads in less than one hour. We use a suite of tumor/germline analysis pipelines for the analysis of DNA-seq data. These are drop-in replacements for GATK best-practices guidelines but highly tuned for faster computational efficiency on the cloud. This reduces computational times for processing raw DNA sequencing data to final somatic mutation calls from order of days to less than two hours. By harnessing these state-of-the-art cloud-computational pipelines, we have substantially improved the turn-around time taken to produce clinical recommendations and analytical reports from patient genomic data.

Approach

This website uses cookies.