In a recent study published in the journal BJC | British Journal of Cancer, researchers explored a novel approach for detecting pancreatic cancer, a disease known for its aggressive nature and poor prognosis.
The researchers combined serum microribonucleic acid or microRNA (miRNA) sequencing with machine learning (ML) techniques to develop accurate diagnostic models for identifying the disease in its often asymptomatic early stages.
a) All the 425 serum samples, including 213 pancreatic cancer patients from 14 hospitals and 212 healthy controls without cancer from three clinics, were randomly divided into the training and validation cohorts. Image Credit: https://www.nature.com/articles/s41416-024-02794-5
Background
Pancreatic cancer is among the most lethal cancers, with a five-year survival rate of only 12%. It is the fourth leading cause of cancer-related deaths worldwide, and its prevalence is expected to rise, potentially becoming the second deadliest cancer by 2030.
Early detection is crucial for improving patient outcomes but remains challenging due to a lack of specific symptoms and reliable biomarkers. Current diagnostic methods, such as imaging and tumor marker analysis like CA19-9, have limitations in sensitivity and specificity, especially for early-stage detection.
miRNAs are small non-coding ribonucleic acid (RNA) molecules that regulate gene expression. They are found in stable forms in bodily fluids, making them potential indicators for various diseases, including cancer.
The advancement in high-throughput sequencing has enabled comprehensive analysis of miRNA expression profiles, providing a powerful tool for identifying potential biomarkers for disease diagnosis and prognosis.
About the Research
In this paper, the authors aimed to develop an innovative technique for early detection of pancreatic cancer by analyzing comprehensive serum miRNA expression profiles using automated ML. Serum samples were collected from 213 pancreatic cancer patients and 212 healthy controls, representing a diverse cohort from 14 hospitals and three clinics.
These samples underwent miRNA sequencing on next-generation sequencing (NGS) platforms, including the NextSeq 550 and Ion GeneStudio S5 systems, to generate a detailed dataset of miRNA expression profiles. Various statistical and ML techniques were then applied to these data to build and validate diagnostic models.
To identify potential biomarkers, the researchers used DataRobot, an automated ML (AutoML) platform, to develop diagnostic models based on the expression profiles of 100 miRNAs and their connection with CA19-9. DataRobot is a powerful tool that automatically generates and evaluates various ML models.
The platform analyzed the miRNA expression data using 64 algorithms, including elastic net and light gradient boosting. Model performance was assessed using a 5-fold cross-validation approach, which repeatedly split the data into training and testing sets to ensure model generalizability.
The study focused on identifying a set of miRNAs that could effectively discriminate between patients with pancreatic cancer and healthy controls. The authors also investigated the potential of combining these miRNAs with the established tumor marker CA19-9 to enhance diagnostic accuracy further.
Research Findings
The outcomes showed that the presented technique accurately detected pancreatic cancer with high sensitivity and specificity. The authors identified 100 highly expressed miRNAs significantly associated with pancreatic cancer.
Combining these miRNAs with CA19-9 achieved an impressive area under the curve (AUC) of 0.99, with a sensitivity of 90% and specificity of 98%, demonstrating the model's strong ability to differentiate between pancreatic cancer patients and healthy controls.
The study also found that the ML model could effectively detect pancreatic cancer in early stages (stage 0-I) with high accuracy. This result is significant since early detection improves treatment outcomes and survival rates.
The model's performance was validated using an independent cohort of 21 asymptomatic patients with pancreatic cancer, further confirming its accuracy and potential for early diagnosis. The models outperformed conventional serum markers like CA19-9, CEA, and DUPAN-2, especially in detecting early-stage disease.
Applications
This research has significant implications for the early detection and treatment of pancreatic cancer. Comprehensive serum miRNA sequencing combined with AutoML provides a non-invasive and highly accurate method for detecting pancreatic cancer, potentially enabling earlier diagnosis and more effective treatment strategies.
This approach could also be applied to other cancer types, allowing for the early detection and treatment of a broader range of malignancies. Additionally, using automated ML platforms like DataRobot can accelerate the discovery of new biomarkers and the development of precise diagnostic tools.
Conclusion
In summary, the proposed approach proved highly effective, accurate, and robust for detecting cancer, particularly pancreatic cancer, at an early stage. It has the potential to transform cancer diagnosis and treatment by serving as a non-invasive screening tool for high-risk groups, such as individuals with a family history of pancreatic cancer or certain genetic syndromes.
Early detection of pancreatic cancer could improve patient health through timely intervention and treatment. Future research should focus on validating these models in larger, prospective studies and assessing their applicability in clinical settings.
Journal Reference
Kawai, M., Fukuda, A., Otomo, R. et al. Early detection of pancreatic cancer by comprehensive serum miRNA sequencing with automated Machine Learning. Br J Cancer (2024). DOI: 10.1038/s41416-024-02794-5, https://www.nature.com/articles/s41416-024-02794-5
Disclaimer: The views expressed here are those of the author expressed in their private capacity and do not necessarily represent the views of AZoM.com Limited T/A AZoNetwork the owner and operator of this website. This disclaimer forms part of the Terms and conditions of use of this website.