Researchers co-led by Ludwig Johns Hopkins’ Christopher Douville described in a Science Translational Medicine paper in January a machine learning approach to profile Alu elements—short, repetitive DNA sequences that jump about the genome and account for about 10% of our genomic sequences—for cancer detection. Alu elements have long had potential as cancer biomarkers but their use for that purpose has been limited by difficulties associated with their analysis that stem from their repetitive nature. Named Alu Profile Learning Using Sequencing (A-PLUS), the method developed by the Ludwig Johns Hopkins team was applied to 7615 plasma samples from 5178 individuals: 2073 of them patients with solid cancers and the remainder controls without a diagnosis of cancer. A-PLUS alone provided a sensitivity of 40.5% across 11 different cancer types in the validation cohort, at a specificity of 98.5%. Combining A-PLUS with DNA sequence-based detection of aneuploidy—an abnormal number of chromosomes common to cancer—and eight common protein biomarkers detected 51% of the cancers at 98.9% specificity. Christopher and his colleagues found that the power of their approach stemmed in large part from a single feature: the global reduction of a subtype of Alu elements—AluS—in the circulating DNA shed by solid tumors. The study suggests the evaluation of Alu elements could improve the performance of several different analytical methods applied to liquid biopsies.
Machine learning to detect the SINEs of cancer
Science Translational Medicine, 2024 January 24