CRIS_single-sample

CRIS_single-sample provides multi-label classification of colorectal cancer (CRC) samples using R to assign multiple CRIS Intrinsic Subtypes (ML²CRIS) and thereby characterize tumor heterogeneity and its associations with prognosis and treatment response.

Key Features:

• Multi-Label Classification: Employs a multi-label approach (ML²CRIS) to assign more than one CRIS Intrinsic Subtype to individual CRC samples, capturing overlapping molecular and phenotypic characteristics.
• Machine Learning Integration: Implements a machine learning-based predictor tailored for single-sample classification to improve accuracy and reliability of subtype assignments.
• Biological Validation: Compares single- and multi-label CRIS associations to validate classifications and demonstrate improved predictive power for prognosis and treatment response.
• Single-Cell Analysis: Incorporates single-cell RNA-seq analysis to elucidate the cellular composition underlying multiple subtype assignments and to distinguish distinct cell populations from hybrid phenotypes.

Scientific Applications:

• Enhancing Prognostic Accuracy: Multi-label CRIS assignments improve prediction of patient prognosis and treatment efficacy.
• Facilitating Personalized Medicine: Detailed subtype profiles support identification of molecular characteristics relevant to therapeutic strategies.
• Expanding Research Horizons: The methodology can be adapted to other cancer types to extend intrinsic subtyping approaches across oncology research.

Methodology:

Developed in R and trained on RNA-seq profiles from 606 CRC patient-derived xenografts (PDXs) together with bulk and single-cell RNA-seq datasets; a multi-label version of the CRIS classifier (ML²CRIS) was developed, validated by comparison to traditional single-label classifications for biological and clinical relevance, and specifically validated for single-sample application.