Firstly, the bioinformatics team from Gene Solutions used whole-genome bisulfite sequencing (WGBS) on five types of tumor tissues (breast, colorectal, gastric, liver and lung cancer) and paired white blood cells (WBC) to construct a tumor-specific methylation atlas (TSMA), where 2,945 CpG regions are discovered between tumor types and WBC. The team then implemented a Deep Learning model of Graph Convolutional Neural Network that combines deconvolution scores from the TSMA with other features to achieve an improved tumor of origin prediction accuracy in the validation dataset of 239 low-depth cfDNA samples.
By enhancing the accuracy of tumor identification through AI-driven method, the study opens up mutliple promises when applying in multi-cancer early detection tests:
Improved Data Accuracy: The combination demonstrated exceptional performance in accurately identifying the tumor of origin, notably up to 100%, 98% and 93% accuracies for breast, liver and colorectal cancer.
Reduced Sequencing Depth: By having a guiding atlas, the R&D team can optimize sequencing depth required for tumor identification, making it more efficient and cost-effective. This reduction in sequencing depth not only accelerates the time-to-result but also conserves costly next-generation sequencing resources.
Optimized Analysis Resources: The AI-driven approach optimizes the use of computational resources, reducing the overall cost and time required for tumor analysis. This optimization is a crucial step towards making advanced circulating tumor DNA analysis accessible and affordable for healthcare providers and patients alike.
Dr. Minh Duy Phan, a lead author of the study commented: “The analytical validation of the new tumor methylation atlas and deep learning algorithm marks a significant milestone in circulating tumor DNA analysis for early cancer signal detection. By harnessing the power of AI, we are enhancing the accuracy and efficiency of multi-cancer early detection technology for real-world utility.”
Future Developments:
With further development by our data team, Gene Solutions is committed to expanding the capabilities of the TSMA. The ongoing research and development efforts aim to enable better quality analysis and application in real-life practice to reduce the cost of assays, ensuring that cutting-edge cancer screening tools are within reach for recommended individuals worldwide.
For more in-depth information about the study and the TSMA, please visit https://spotmas.com/blog/pioneering-a-tumor-specific-methylation-atlas-tsma-to-identify-tissue-of-origin-too-in-multi-cancer-early-detection/
References:
Nguyen, T.H., Doan, N.N.T., Tran, T.H. et al. Tissue of origin detection for cancer tumor using low-depth cfDNA samples through combination of tumor-specific methylation atlas and genome-wide methylation density in graph convolutional neural networks. J Transl Med 22, 618 (2024), doi: 10.1186/s12967-024-05416-z
