Commercial ctDNA tests are now widely used to identify patients for biomarker-directed therapies (e.g. PARP inhibitors in BRCA2-defective cancers) but are uninformative when ctDNA fraction is low. We have developed this research tool to help users decide on whether to pursue ctDNA- or tissue-based biomarker testing, by predicting a patient’s ctDNA fraction based on routinely collected clinical information.
Enter in values for as many of the clinical variables as you have available. All fields are optional but the more you provide the more accurate the results will be. Press "predict ctDNA%", and the tool will predict the patient's ctDNA fraction, as well as the probability of the patient having a ctDNA fraction greater than 2%.
All humans carry short circulating DNA fragments, known as cell-free DNA (cfDNA), in their blood. These DNA fragments originate from damaged or dying cells in the body. In patients with cancer, a significant fraction of cell-free DNA in the blood can originate from tumor cells. These tumor-derived DNA fragments are referred to as circulating tumor DNA (ctDNA). By capturing and reading the circulating tumor DNA present in a patient's blood, clinicians and oncologists can characterize the DNA alterations that drive the patient's cancer.
The amount of circulating tumor DNA in a patient's bloodstream depends on the number of cancer cells in their body. It is often quantified as the percentage of circulating tumor DNA, relative to total cell-free DNA in the sample. This percentage is referred to as the circulating tumor DNA fraction (ctDNA fraction, ctDNA%). If the circulating tumor DNA fraction is very low, it may be impossible to detect cancer driving mutations in the blood sample, and a more costly tissue biopsy may be required instead.
Predictions are based on a K-nearest neighbor model trained using 472 cell-free DNA samples from metastatic prostate cancer patients accrued to our liquid biopsy collection program in Vancouver, BC, Canada. To make it easier to evaluate the reliability of prediction results, each prediction is accompanied by a table showing the clinical data and ctDNA fractions for 10 patient records that most closely match the queried clinical attributes. If more than 10 patient records match the query equally well, the ctDNA fraction is estimated based on all of those records.
Predicted ctDNA fractions are valid for metastatic castration-resistant prostate cancer patients with progressive disease. This means patients that are not receiving systemic therapy or whose cancer is progressing despite ongoing treatment. Blood samples obtained during effective treatment often contain significantly lower ctDNA fractions.
Cell-free DNA concentration, lactate dehydrogenase (LDH), alkaline phosphatase (ALP), and presence of visceral metastases are most strongly correlated with ctDNA fraction. See the figure below for more details:
This website is the product of a collaboration between the Wyatt Lab at the Vancouver Prostate Centre (Canada) and the Computational Biology Group at Tampere University (Finland). Methods development and data analysis was performed by Matti Annala, Wilson Tu, Cameron Herberts, Nicolette Fonseca, and Sinja Taavitsainen. Clinical data was collected by Daniel Khalaf, Corinne Maurice-Dror, and Edmond Kwan. A manuscript about the ctDNA fraction prediction methodology has been submitted for review.