PRISM

PRISM performs deconvolution of bulk RNA sequencing data using single-cell reference datasets to extract sample composition and cell-type-specific whole-transcriptome profiles for analysis of heterogeneous and evolving cancer transcriptomes.

Key Features:

• Simultaneous Extraction: Simultaneously estimates sample composition and cell-type-specific whole-transcriptome profiles from individual bulk RNA-seq samples.
• Adaptation to Individual Samples: Adapts analyses to each sample to account for systematic composition changes due to chemotherapy and anatomical site variation.
• Composition-Free Transcriptomic Profiles: Produces composition-free transcriptomic profiles that remove confounding effects of varying sample composition and improve prediction of patient treatment responses relative to composite bulk analyses.
• Validation and Verification: Validated in independent cohorts including ovarian cancer and melanoma, with composition estimates and cell-type-specific expression confirmed by whole-genome sequencing and RNA in situ hybridization.

Scientific Applications:

• Cancer Research: Separates and analyzes cancer cells and the tumor microenvironment to study tumor heterogeneity, progression, and treatment response.
• Personalized Medicine: Provides cell-type-resolved transcriptomic profiles to support prediction of patient-specific responses to therapies.
• Longitudinal Studies: Tracks changes in tumor composition and cell-type-specific expression over time, including responses to chemotherapy.

Methodology:

Poisson-based latent statistical framework that integrates single-cell reference data with bulk RNA-seq to deconvolve complex transcriptomic data into sample composition and cell-type-specific expression.