On enabling early detection of cancer via DNA-based signatures in liquid biopsy.

Abstract

e24262 Background: Early detection of cancer has been in the spotlight, due to its promise to radically change the course of treatment of this disease. If cancer is detected at early stages, the probability of having positive outcome is significantly higher than detecting it at late stages, i.e., the current state-of-the-art. While circulating tumor cells (CTCs) provide a viable method for liquid biopsy, a major challenge with them for earlier stages of cancer has been their low sensitivity. However, with the technological advances made in this field, higher sensitivities are now reported to be achievable. Once the CTCs are isolated, if the cells are sequenced and specific cancer signatures are found, then the problem of early detection could become addressable. Methods: The focus of this work is on finding sensitive and specific signatures for identification of cancer. Previously, we reported on a molecular signature for early detection of breast cancer, based on whole genome sequencing (WGS) data. Here, we report on different WGS-based signatures, tested on a small representative set of real-patient, tissue-based data from the International Cancer Genome Consortium (ICGC). Results: Our preliminary results suggest that a sensitivity rate of ~90% would be possible as the upper limit (i.e., not including the sensitivity of the CTC detection) for the detection of cancer at Stages I or II, for a limited number of cancer types, e.g., breast, prostate and pancreatic cancer. While these signatures were found using tissue-based data, they are expected to remain valid when applied to CTCs, despite the sources of noise that are attributed to CTCs, namely allele dropout (ADO) and additional false positives/negatives. The observations were verified on a few late-stage CTCs --to be repeated once early-stage CTCs are available. Conclusions: In summary, this preliminary study supports the claim that a CTC-based liquid biopsy technique using WGS has the promise to provide a reliable platform for early detection of cancer. In addition to the isolated signatures found in this study and the previous study, the combination of these signatures via machine learning techniques could result in a yet more sensitive and specific early predictor for a subset of cancer types.