CPSM: an R package for cancer patient survival risk model using transcriptomics and clinical data.

Area of Science:

• Computational biology
• Bioinformatics
• Cancer research

Background:

• Traditional survival analysis methods like Kaplan-Meier curves lack individual patient detail.
• Multi-omics data in cancer presents challenges for personalized survival modeling due to complexity and high dimensionality.
• There is a need for integrated, reproducible tools to generate individualized survival predictions.

Purpose of the Study:

• To introduce the Cancer Patient Survival Model (CPSM) R package.
• To provide an end-to-end computational pipeline for personalized survival and risk predictions using multi-omics and clinical data.
• To enhance patient-clinician communication through interpretable, individualized survival visualizations.

Main Methods:

• Developed CPSM R package with 10 functions covering data preprocessing, feature selection, survival risk modeling, and visualization.
• Utilized publicly available TCGA datasets for glioblastoma multiforme, acute myeloid leukemia, pancreatic adenocarcinoma, and breast invasive cancer.
• Employed repeated cross-validation with uncertainty quantification for robust model evaluation in high-dimensional, small-sample settings.

Main Results:

• CPSM efficiently handles high-dimensional datasets (e.g., >60,000 RNA transcripts) and diverse clinical variables.
• The package enables robust and interpretable individualized survival predictions across different data conditions.
• Demonstrated utility across four distinct cancer types, showcasing versatility.

Conclusions:

• CPSM offers an efficient, user-friendly, end-to-end solution for personalized cancer survival prediction.
• Integrated visual tools within CPSM improve interpretability and support informed clinical decision-making.
• The R package is freely available on Bioconductor and GitHub for broader accessibility.