Machine learning for early cancer detection
A study perfected a machine learning-based test, the A-PLUS, by using the differentiated presence of Alu elements in cancer cells; helping early cancer detection.
The discovery of Alu elements as biomarkers
An Alu sequence is a Short INterspersed Element (SINE) of approximately 300 base pairs. It is estimated that there are more than one million Alu sequences in the human genome, making up 10% of the total human genome. The insertion of Alu sequences is implicated in several human hereditary diseases and various forms of cancer.
Alu elements, numerically limited compared to the billions of elements that make up DNA, have long been neglected as biomarkers due to their complex and repetitive nature. However, research conducted by Douville et al. has opened up new perspectives. The study involved 11 cancer types and over 7,600 blood samples, showing that the reduction of AluS subfamily elements is particularly characteristic of solid tumours.
The A-PLUS test: a step forward in cancer diagnosis
Douville and colleagues collected samples from 3,105 people with solid tumours and 2,073 without. The study covered 11 cancer types and 7,615 blood samples. The replicates were used to test how well the model worked.
The A-PLUS test, combining Alu element analysis with aneuploidy and common protein biomarkers, achieved a sensitivity of 40.5% on 11 different cancer types, with a specificity of 98.9%. This level of specificity is crucial to avoid false positives in screening tests, ensuring accurate and timely diagnosis.
The method was designed to achieve high specificity in cancer classification and was validated in several independent cohorts, with solid tumours particularly characterised by a reduction in AluS subfamily elements. The discovery of Alu elements as a key component in cancer diagnosis offers a new complementary approach to current diagnostic methodologies for detecting tumours at the earliest stage.
Indeed, 99% of people diagnosed with stage 1 breast cancer will be alive five years later; however, if the cancer is detected at stage 4, when the disease has spread to other organs, five-year survival drops to 31%.
Looking into the future
Despite constituting only 11% of human DNA, Alu elements prove invaluable for the early detection of cancer. The reduction of AluS elements in patients with solid cancer is a hallmark that could further improve the effectiveness of multiple diagnostic methodologies. The next step will be the selection and optimal combination of biomarkers for an even more precise and personalised diagnosis.
In conclusion, the research paves the way for a new era in cancer diagnosis, in which Alu elements in blood turn out to be not only DNA repeats, but true indicators of health status.
- Douville C, Lahouel K, Kuo A, Grant H, Avigdor BE, Curtis SD, Summers M, Cohen JD, Wang Y, Mattox A, Dudley J, Dobbyn L, Popoli M, Ptak J, Nehme N, Silliman N, Blair C, Romans K, Thoburn C, Gizzi J, Schoen RE, Tie J, Gibbs P, Ho-Pham LT, Tran BNH, Tran TS, Nguyen TV, Goggins M, Wolfgang CL, Wang TL, Shih IM, Lennon AM, Hruban RH, Bettegowda C, Kinzler KW, Papadopoulos N, Vogelstein B, Tomasetti C. Machine learning to detect the SINEs of cancer. Sci Transl Med. 2024 Jan 24;16(731):eadi3883. doi: 10.1126/scitranslmed.adi3883. Epub 2024 Jan 24. PMID: 38266106.