MnM is an R-package that includes the code for running the newly developed machine-learning algorithm M&M, as short for Minority & Majority classifier. This classifier enables the classification of pediatric tumor (sub)types based on RNA-seq data. Besides the creation of the classifier models, MnM contains many functions to visualize M&M's performance on reference and test cohorts.
With over 120 documented tumor types and 250 tumor subtypes, identifying the correct tumor entity during the diagnostic procedure remains a challenging but crucial process within pediatric oncology. M&M allows for inclusion of many rare pediatric tumors with as few as three available samples, occurring with rates of less than once for every 500 children diagnosed with cancer. M&M is designed to assist pathologists during the diagnostic process to reduce inter-observer variability and help recognizing rare pediatric tumors.
library(remotes)
remotes::install_github("princessmaximacenter/MnM/", dependencies = T)
Please see the supplied tutorial, vignettes, and documentation within R on how to properly use all functions. If you wish to familiarize yourself with MnM, the tutorial is the best starting point (DataTutorial). For this purpose, you can download the ZIP of this github page (click <> Code, Download ZIP).
RNA TPM-normalized count data and metadata of the reference cohort and test cohort can be obtained from ArrayExpress (accession E-MTAB-14038). The final resulting pre-trained models, which can be used to classify new incoming RNA-samples, can be obtained from Zenodo (https://zenodo.org/records/11575098). These models now are also capable of performing missing gene imputation, making them more user-friendly. Please note that RNA-transcript count rownames are required to be HGNC-symbols, ENSEMBL-IDs currently cannot be used.
In case of questions, suggestions or additional comments, please reach out to us via p.kemmeren[AT]prinsesmaximacentrum.nl (orcid ID: 0000-0003-2237-7354).