Identification of Cholangiocarcinoma (CCA) Subtype-Specific Biomarkers

Abstract

Liver cancer ranks sixth globally in diagnoses and second in cancer-related deaths. Cholangiocarcinoma (CCA), a relatively rare cancer originating from bile duct epithelium, constitutes 2% of all cancers, with increasing occurrences in Westerns. Incidence is influenced by inflammation, genetics, risk factors, and regional disparities, with higher rates in the Eastern hemisphere. Diagnostic and prognostic biomarkers are pivotal for effective cancer prevention and management. Recent research explores serum proteins for non-invasive CCA diagnosis and proposes targeted receptor approaches for therapeutics.This study aims to identify these biomarkers via bioinformatics analysis of public datasets, focusing on CCA patient transcriptomes to uncover gene biomarkers linked to age and survival. Pathway analysis reveals functions and pathways associated with these biomarkers. Additionally, the study employs the ESM-TFpredict machine learning model to predict transcription factors (TFs) using protein sequence data.Leveraging publicly available data enhances our understanding of liver cancer’s molecular profiles and clinical relevance, particularly concerning CCA, This study integrates bioinformatics analysis, transcriptomic exploration, and machine learning to unveil a novel set of potential diagnostic and prognostic biomarkers for CCA.

Novel hepatocellular carcinomas (HCC) Subtype-Specific Biomarkers

“…The degree of difficulty for predicting transcription factors (TFs) stems from the intricate and resource-intensive process of experimental validation [66][67][68]. In this research, we utilized our developed ESM-TFpredict framework, designed to automate the identification of potential TFs [58,69]. To enhance training efficiency while working with limited resources, the sequence is divided into predefined windows.…”

“…Using our devised ML/DL-based ESM-TFpredict model [58,69], we have effectively identified 102 transcription factors (TFs) out of the 647 HCC-specific diagnostic and prognostic biomarkers that we previously uncovered (Table 1, Supplementary Table 1). Our findings align with transcription profiling data observed in major cancer types, especially HCC [34,36,38,39,41,72].…”

“…Pathway analysis helps elucidate functions and disrupted pathways correlated with these biomarkers. The study also explores the prediction of transcription factors (TFs) using a machine learning model called ESM-TFpredict [58,69], which lead to the identification of 102 potential TFs, leveraging protein sequence data. The dataset preprocessing involves curating TF sequences and non-TF sequences for analysis.…”

“…Pathway analysis illuminated functions and disrupted pathways related to these biomarkers, providing insights into disease progression. Additionally, altered CCM protein levels in tumors propose their potential as biomarkers spanning multiple cancers [20,[53][54][55][56][57][58][59][60][61][62][63]. Both progesterone receptor types (nPRs and mPRs) could potentially serve as biomarkers for genetically influenced cancer subtypes [20,[61][62][63][64][65].…”