PRISM

PRISM infers the composition of epigenetically distinct tumor subclones from reduced representation bisulfite sequencing (RRBS) DNA methylation patterns to characterize tumor heterogeneity and global DNA methylation reprogramming.


Key Features:

  • Methylation pattern-based inference: Utilizes subclonally reprogrammed methylation patterns and dichotomous methylation states (fully methylated vs. unmethylated) within short genomic regions to reveal constituent tumor populations.
  • Reference-free analysis: Performs inference without reliance on a reference genome, using methylation pattern signals alone.
  • RRBS and Bismark compatibility: Designed for RRBS data and requires mapping results from Bismark (BAM files) as input.
  • HMM-based error correction: Incorporates an in silico proofreading step using a DNA methyltransferase 1-like hidden Markov model to correct erroneous methylation patterns.
  • Beta-binomial mixture modeling: Models frequencies of dichotomous methylation patterns using a beta-binomial mixture model to obtain sufficient statistics for subclonal abundance.
  • Expectation-maximization fitting: Fits the statistical model with an expectation-maximization algorithm to infer subclonal composition and evolutionary relationships.

Scientific Applications:

  • Tumor subclone deconvolution: Infers relative abundances of epigenetically distinct subclones from RRBS-derived methylation patterns.
  • Study of epigenetic reprogramming: Characterizes global DNA methylation reprogramming in cancers, including myeloid malignancies.
  • Epigenetic heterogeneity and evolution: Reconstructs aspects of tumor heterogeneity and evolutionary history from methylation pattern data.

Methodology:

Inputs are BAM files of RRBS reads aligned by Bismark; methylation patterns are identified and corrected using a DNA methyltransferase 1-like hidden Markov model; frequencies of dichotomous methylation patterns in short regions are modeled with a beta-binomial mixture model; the model is fitted using an expectation-maximization algorithm to infer subclonal composition.

Topics

Details

License:
MIT
Programming Languages:
Python
Added:
11/14/2019
Last Updated:
12/6/2020

Operations

Publications

Lee D, Lee S, Kim S. PRISM: methylation pattern-based, reference-free inference of subclonal makeup. Bioinformatics. 2019;35(14):i520-i529. doi:10.1093/bioinformatics/btz327. PMID:31510697. PMCID:PMC6612819.

PMID: 31510697
PMCID: PMC6612819
Funding: - Ministry of Science: No.NRF-2017M3C4A7065887 - ICT: No.NRF2014M3C9A3063541 - Ministry of Health & Welfare, Republic of Korea: HI15C3224