CancerRadar is a blood test, which is highly sensitive in detecting a trace amount of tumor DNA in cell-free DNAs.
CancerRadar combines our experimental and computational innovations: (1) we developed an experimental method to cost-effectively sequence cell-free DNA methylomes; and (2) our computational platform extracts and combines multiple types of features (methylation, copy number variation, fragment length properties, microbial abundances) from the cell-free DNA methylome data, and conducts ensemble machine learning for high-performance cancer detection.
Why is CancerRadar so powerful for cancer detection?
CancerRadar is highly sensitive in detecting a trace amount of tumor DNA in cell-free DNAs.
While many panel-based approaches focus on capturing tumor DNAs only from a few genomic regions, our methylome profiling examines millions of genomic regions in an unbiased manner, therefore more robustly capturing tumor DNA fragments.
CancerRadar simultaneously searches for genetic, epigenetic, and even microbial signals from tumors.
Compared to the commonly used approach that focuses on one type of tumor signals, CancerRadar can combine multiple types of signals, therefore sensitively detecting tumors.
CancerRadar can continuously learn and adapt according to our rapidly growing database of training samples
Given the diversity of cancer (types, subtypes, etiologies) and the heterogeneity of the patient population (age, gender, ethnicity, and comorbidity), the currently available sample sizes are small even in the largest clinical studies to date. The CancerRadar system cost-effectively retains the genome-wide epigenetic and genetic profiles of cancer abnormalities, thereby permitting classification models to learn newly significant features as training cohorts grow, as well as expanding their scope to more cancer types. Therefore, CancerRadar can truly facilitate a big data approach for cancer detection.