yingwang07 / ptem

A two-step approach to Predict Tumor Evolution Model

PTEM

A two-step approach to Predict Tumor Evolution Model

Description

This is an apporch to predict tumor evolution model using single cell copy number profile. It will estiamted:

• It the tumo cell population follows neutral or selective evolution model.
• If the tumor cell popoulation follows selective eovlution model, it is linear, branching or punctuated eovlution model.

System requirements and dependency

This approach runs on R (version > 4.0) and has dependency on the R packages: diptest.

Usage

Please download and copy the distribution to your specific location.

  setwd("./PTEM")
  source("model.function.R")

Input a CNV frequncey vector calculated from a single cell coy number profile. Each element in the vector corresponds to the frequency of genome region/gene gain or loss in the tumo cell population.

Generate a CNV frequency vector from neutral evolution model i.e. unimodal distribution

 CNVfre1 <- rnorm(500,mean=0.3,sd=0.02)

Generate a CNV frequency vector from selective evolution model i.e. multimodal distribution

 CNVfre2 <- c(rnorm(100,mean=0.1,sd=0.02),rnorm(200,mean=0.4,sd=0.02),rnorm(200,mean=0.6,sd=0.02))

Estimating if the CNV frequency vector follows neutral evolution

  modalityEst1 <- dip.test(CNVfre1)
  modalityEst2 <- dip.test(CNVfre2)

For scDNA-seq, keep CNV frequency > 0.05; For scRNA-seq, keep CNV frequenct > 0.02

Neutral evolution model

modalityEst1

Hartigans' dip test for unimodality / multimodality

data:  CNVfre1
D = 0.013069, p-value = 0.8705
alternative hypothesis: non-unimodal, i.e., at least bimodal

Selective evolution model

modalityEst2

Hartigans' dip test for unimodality / multimodality

data:  CNVfre2
D = 0.11944, p-value < 2.2e-16
alternative hypothesis: non-unimodal, i.e., at least bimodal

For selectiv evolution model, further distinguish linea, branching and punctuated evolution model

cds <- selectModel(CNVfre2,cutoff1=0.01,cutoff2=0.7)
modelEst <- ModelPredict(cds)

modelEst

  branch       linear punctuated    Predict
1 0.01829491 5.971049e-09  0.9817051 Punctuated

The first to the third column are the posterior probabilities of corresponding selective eovlution model. The Predict column returns the model with the largest posterior probability.