Reliable detection of cancer at an early stage before symptoms appear has the potential to dramatically decrease global cancer mortality.
The GRAIL approach
GRAIL is poised to detect cancer early in asymptomatic individuals by combining high-intensity sequencing of unprecedented breadth and depth with the techniques of modern data science. Through one of the largest clinical trial programs ever pursued in genomic medicine, GRAIL will create vast datasets to develop our products and demonstrate their clinical utility.
Look deeper and broader
Through mechanisms that are not well-understood, small fragments of tumor DNA appear in systemic circulation. An increasing body of evidence demonstrates that these fragments, known as circulating tumor DNA (ctDNA), directly encode genetic features of the tumors from which they are derived.
Initiatives and tools elucidating the genomic and molecular biology of tumors have greatly enhanced our understanding of cancer. The Cancer Genome Atlas (TCGA) project, for example, has established that cancers are extremely heterogeneous.
The complexity of tumors suggests that measurements that capture the breadth of the heterogeneous cancer genome represent the best and most direct signature of disease biology. GRAIL’s unique ability to sequence broadly across the genome allows us to discover the millions of unique patterns that define cancer.
Because ctDNA is a more direct measure of cancer and represents the genomic diversity of cancer throughout the body, ctDNA-based tests have the potential to deliver unparalleled sensitivity and specificity for early cancer detection.
The fraction of ctDNA in the bloodstream compared to DNA from non-cancerous cells is very small, necessitating the ability to distinguish faint signals of early tumors from an overwhelming background of genomic DNA.
GRAIL’s ability to query ctDNA at extraordinary depth allows us to detect those faint ctDNA signals in a sea of background noise.
The scale of GRAIL’s
From the laboratory to the clinic, GRAIL aims to produce the highest-quality data and transform it into clinically actionable insights. Our high-intensity sequencing assays will yield approximately one terabyte of data for every individual, creating datasets of unprecedented scale in modern clinical medicine.
We are deploying, at scale, the latest tools of data science, including powerful approaches from machine learning such as hierarchical neural networks. We intend to apply such methods to the ultimate problem of classifying patients according to the presence, type, and severity of cancer, and in all steps of our data-generating pipeline.
Our commitment to scientific rigor and clinical utility
GRAIL is committed to developing a deep understanding of cancer biology at the foundation of our measurements. We must also demonstrate unambiguously that the tests we develop with that knowledge will improve patient outcomes when applied in a real population.
To do either requires us to conduct clinical studies and trials that will enroll tens of thousands of cancer patients and healthy individuals. GRAIL thus plans to conduct some of the largest cancer-related studies to date to generate scientific and clinical evidence of our impact.