TRANSCUP

TRANSCUP predicts the primary site of cancer of unknown primary (CUP) from next-generation RNA sequencing (RNA-seq) transcriptome data using machine learning classification.

Key Features:

• Raw Data Processing: Handles raw next-generation RNA-seq data to prepare inputs for downstream analysis.
• Read Mapping: Maps sequencing reads to a reference genome to generate alignments for expression analysis.
• Quality Report Generation: Produces comprehensive quality reports assessing integrity and reliability of processed RNA-seq data.
• Gene Expression Quantification: Quantifies gene expression levels from RNA-seq data for transcriptome profiling.
• Machine Learning Model Building: Builds tumor type classifiers using a random forest algorithm.
• Training on External RNA-seq Datasets: Trains models using external RNA-seq datasets to improve classification performance.
• Scalable Workflow: Integrates processing, mapping, QC, quantification, and modeling into a scalable computational workflow for large datasets.
• Performance Metrics: Demonstrates high accuracy, sensitivity, and specificity in tumor type classification.

Scientific Applications:

• CUP Primary Site Prediction: Classifies the tissue or organ of origin for cancers of unknown primary using transcriptome signatures.
• Tumor Type Classification: Distinguishes between tumor types based on RNA-seq gene expression profiles.
• Clinical Decision Support: Provides molecular classification information that can inform selection of targeted treatment strategies.

Methodology:

Computational steps include raw RNA-seq data processing, read mapping to a reference genome, quality report generation, gene expression quantification, training random forest models on external RNA-seq datasets, and tumor type classification.