Abstract A067: lncTransformer: Identifying novel prognostic biomarkers for pancreatic cancer using deep learning models

Abstract Pancreatic cancer remains a formidable challenge due to high mortality rates and limited therapeutic options due to late detection of the cancer. Recent literature has focused on investigating the potential of long noncoding RNAs (lncRNAs) as biomarkers in several cancers. To identify significantly altered expression of lncRNA in pancreatic cancer, I collected information from the Cancer Genome Atlas (TCGA) and extracted RNA-sequencing (RNA-seq) transcriptomic profiles of pancreatic carcinomas. Out of 60,660 gene transcripts shared between 151 pancreatic cancer patients, 38 lncRNAs were identified to be significantly differentially expressed. To find the lncRNAs related to metastatic progression of pancreatic cancer, different machine learning algorithms, including logistic regression (LR), support vector machine (SVM), random forest classifier (RFC), and a custom autoencoder model were trained to identify potential prognostic biomarkers. The custom deep learning model predicted metastatic prognostic biomarkers with a 92% accuracy, with the second-best accuracies coming from the SVM and RFC models, having an accuracy of 76%. Using the deep learning model, the following novel lncRNA and gene biomarkers were identified: LINC01300, SERPINB13, AC010789.1, TMPRSS15, DUSP5-DT, AL513128.3, MIR205HG, LINC00486, RF00019, LINC01115, and AC133530.1. Further investigating these results, specifically the identified lncRNA, the LINC01300 was identified to be potentially circulating in the blood. If further analysis proves this hypothesis correct then LINC01300 along with other identified genes could serve as a biomarker panel for a liquid biopsy. The identified biomarkers would be especially helpful in rural areas with lower accessibility to healthcare as it would allow for an easier, more efficient, and cost-effective method of pancreatic cancer detection. Based on these findings, further investigations of this gene panel are being conducted, in vitro and in vivo, to create a novel liquid biopsy to assist in the prognostic diagnosis of metastatic progression of pancreatic cancer. Citation Format: Shivali Singh. lncTransformer: Identifying novel prognostic biomarkers for pancreatic cancer using deep learning models [abstract]. In: Proceedings of the AACR Special Conference: Liquid Biopsy: From Discovery to Clinical Implementation; 2024 Nov 13-16; San Diego, CA. Philadelphia (PA): AACR; Clin Cancer Res 2024;30(21_Suppl):Abstract nr A067.