Deciphering the Diagnostic Potential of Small Non-Coding RNAs for the Detection of Pancreatic Ductal Adenocarcinoma through Liquid Biopsies
​
Hadas, Volkov, Tel Aviv University Edmond J. Safra Center for Bioinformatics.
Rani, Shlayem, Tel Aviv University. Maria, Raitses-Gurevich, Cancer Center Sheba Medical Center Tel Hashomer.
Noam, Shomron, Tel Aviv University Edmond J. Safra Center for Bioinformatics.
Background & Objective: Pancreatic Ductal Adenocarcinoma (PDAC) is highly lethal due to lack of high sensitivity diagnostic methods. This study identifies an ensemble of unique circulating cell-free non-coding RNA (ncRNA) signatures for PDAC detection through advanced sequencing, bioinformatics, and machine learning (ML).
Methods: We analyzed plasma samples from control and PDAC patients. An innovative stepwise mapping strategy and bioinformatics pipeline were developed to quantify diverse ncRNA species, including microRNAs (miRNAs), long non-coding RNAs (lncRNAs), and other small RNA subtypes. Differential expression analysis identified ncRNAs significantly altered between patients and controls. ML algorithms ranked gene importance to select key ncRNAs, and a classifier was trained using the top-segregating ncRNAs to evaluate diagnostic accuracy. Pathway analyses validated the biomarkers' biological relevance to PDAC.
Results: Twenty ncRNAs were identified as significant discriminators. Our integrated bioinformatics and ML approach achieved a classification accuracy of 87% and an area under the receiver operating characteristic curve of 91%, surpassing existing ncRNA based diagnostic methods. Top ncRNAs encompassed various classes, including multiple microRNAs, lncRNAs, and Y_RNA, each exhibiting distinct expression patterns associated with PDAC. Pathway analyses revealed that these ncRNAs are involved in critical pancreatic cancer-related pathways, such as cell proliferation, MAPK signaling and cellular senescence, underscoring their biological significance and potential roles in PDAC.
Conclusion: Combined analysis of multiple ncRNA subtypes using integrated bioinformatic and ML methods presents an advancement in the non-invasive diagnosis of PDAC. Unlike traditional methods that focus on a single RNA subtype, our ensemble ncRNA strategy leverages collective diagnostic potential. We identified key ncRNA signatures that are intricately involved in PDAC-specific pathways. This comprehensive approach enhances detection capabilities and provides biological evidence into PDAC pathogenesis. Our work demonstrates the effectiveness of combining multi-class ncRNA with ML to improve diagnostic performance, and for the development of more effective non-invasive diagnostics.