Technology ID

Personalized Cancer Evaluation (PERCEVAL) Method and Software

Lead Inventor
Campo, David (CDC)
Software / Apps
Development Stages
Pre-Clinical (in vitro)
Research Products
Computational models/software
Lead IC
Cancer represents the leading cause of morbidity and mortality worldwide, with approximately 14 million new cases and 8.2 million cancer related deaths in 2012. This number is predicted to rise by approximately 70% over the next two decades according to the World Health Organization. Prognosis depends heavily on both early detection and frequent monitoring of the patient's response to treatment. Cancerous tumors shed nucleic acids into blood, which can be detected by ultra-deep sequencing of mitochondrial DNA (mtDNA). This method has the potential to provide early detection in asymptomatic individuals or those with vague, undefined symptoms. Currently, researchers are identifying specific changes in the whole human genome that occur in tumors of patients. These targeted sequences are then compared to other individuals.

Unfortunately, cancer mutations are often unique to a patient or are uncommon across patients with similar cancers. CDC researchers developed PERCEVAL, a software that is capable of detecting cancer in asymptomatic individuals based on mitochondrial DNA (mtDNA). In addition to early detection, PERCEVAL has the potential to determine the severity of the cancer and monitor whether a given therapy is working for an individual patient.

Commercial Applications
  • Method and software with initial accuracy in detecting liver cancer in samples
  • Method and software for early detection of universal cancers
  • Method and software for determining cancer severity and tumor stages of patients already diagnosed
  • Method and software to monitor if cancer treatment is working
Competitive Advantages
  • The model correctly classified 232 Liver Cancer and 232 Non-cancer samples with accuracy of 99.78% and average accuracy of 92.23% in 10-fold cross-validation. In further validation, the model accurately separated 93.08% of Liver cancer samples (n=61) and Non-cancer samples (n=159) that were not previously seen by the model.
  • For liver cancer, detection method does not rely on biomarkers that can result in false positives (e.g., for acetaminophen or alcohol abuse)
  • Non-invasive method
  • May be capable of cancer detection before the patient exhibits symptoms
  • System runs on Linux and applies machine learning algorithms to create a model that better generalizes prediction to new samples
  • Uses a single region, mtDNA, circulating in blood which has been shown to mutate in a wide array of cancers
Licensing Contact:
Mitzelfelt, Jeremiah