RaMBat : A gene-Ranking based framework for identifying Medulloblasotma (MB) subtypes with severe batch effects

Publication

Accurate identification of medulloblastoma subtypes from diverse data sources with severe batch effects by RaMBat. Mengtao Sun, Jieqiong Wang and Shibiao Wan*. BioRxiv 2025.02.24.640010; DOI: https://doi.org/10.1101/2025.02.24.640010

Workflow

Installation

To install the package in RStudion:

devtools::install_github('wan-mlab/RaMBat')
library(RaMBat)

Train your own data to extract features

1. use GRA() function to extract differentially ranked genes (gene rank analysis)
all_rank_t_genes<- GRA(data, sampAnnot)

2. use RRA() function to extract gene ratios (GERs) (Reversal ratio analysis)
all_reversed_gp_genes<-RRA(data, sampAnnot, all_rank_t_genes)

3. use LaSelect() function to select non-zero coefficient GERs (LASSO feature selection)
MB_RANK_GP<-LaSelect(data, sampAnnot, all_rank_t_genes,all_reversed_gp_genes)

4. use predMB() function to predict the MB subtype information (MB subtype identification)
Predicted_subtype<- predMB(test_data,MB_RANK_GP)

Predict the MB subtype information directly

data(MB_RANK_GP)
Predicted_subtype<-predMB(test_data,MB_RANK_GP)

Example

1. Extrat features from your own training dataset (training_dataset), then predict the subtype information for the new test MB samples (new_MB_samples). in this part, a samp_Annot file need to be prepared by the user (check the format of 'sampAnnot_GSE85217' in data folder).

1. all_rank_t_genes<- GRA(training_dataset, samp_Annot)
2. all_reversed_gp_genes<-RRA(training_dataset, samp_Annot, all_rank_t_genes)
3. MB_RANK_GP<-LaSelect(training_dataset, samp_Annot, all_rank_t_genes,all_reversed_gp_genes)
4. myMat<- predMB(new_MB_samples,MB_RANK_GP)

2. Predict the subtype information for given new test MB sample (new_MB_samples) based on features selected by RaMBat from dataset GSE85217.

data(MB_RANK_GP)
myMat<-predMB(new_MB_samples, MB_RANK_GP)

Evaluate the performance of RaMBat

Load the necessary data. all_13datasets is the normalized combination of all 13 independent test datasets. samp_13 is the annotation file for all_13datasets

data(MB_RANK_GP)
data(samp_13)
data(all_13datasets)
Pred_performance<-testMB(all_13datasets, MB_RANK_GP, samp_13)