Leveraging sequences missing from the human genome to diagnose cancer.

Background: Cancer diagnosis using cell-free DNA (cfDNA) has the potential to improve treatment and survival but has several technical limitations.

Methods: In this study, we developed a prediction model based on neomers, DNA sequences 13-17 nucleotides in length that are predominantly absent from the genomes of healthy individuals and are created by tumor-associated mutations.

Results: We show that neomer-based classifiers can accurately detect cancer, including early stages, and distinguish subtypes and features. Analysis of 2577 cancer genomes from 21 cancer types shows that neomers can distinguish tumor types with higher accuracy than state-of-the-art methods. Generation and analysis of 465 cfDNA whole-genome sequences demonstrates that neomers can precisely detect lung and ovarian cancer, including early stages, with an area under the curve ranging from 0.89 to 0.94. By testing various promoters or over 9000 candidate enhancer sequences with massively parallel reporter assays, we show that neomers can identify cancer-associated mutations that alter regulatory activity.

Conclusions: Combined, our results identify a sensitive, specific, and simple cancer diagnostic tool that can also identify cancer-associated mutations in gene regulatory elements.