inst/Methods.md

In jasonserviss/mesenchymalSubsetAnalysis:

Mesenchymal subtype analysis

Classification Cell types identified as mesenchymal via their expression of the THY1 marker were sub-classified using the Mclust1 software version 5.3 within the R statistical programming language2. Fourteen Gaussian finite mixture models with component numbers ranging from 1-20 were applied and evaluated via the Bayesian information criterion. A spherical, varying volume model with 10 components was found to best represent the underlying data. The uncertainty in converting a conditional probablility from expectation–maximization to a classification in model-based clustering was reported and cells with a uncertainty > 0.2 (105 cells = ~16%) were removed from further downstream analysis.

Gene expression profile identification Identified mesenchymal subtypes were grouped with previously characterized major cell types and analysed to identify gene expression profiles. Approximately 6000 genes were pre-selected by choosing those 200 genes exhibiting the highest absolute fold change for each cell type classification. Next, the Kolmogorov-Smirnov test was used to compare each pairwise cell type class for each of the pre-selected genes. 583 genes were identified as significantly (p < 0.05) differentially expressed in all pairwise tests for a single class and were annotated as being specific for that class.

1 Scrucca L., Fop M., Murphy T. B. and Raftery A. E. (2016) mclust 5: clustering, classification and density estimation using Gaussian finite mixture models The R Journal 8/1, pp. 205-233 2 R Core Team. R: A Language and Environment for Statistical Computing, https://www.R-project.org/ (2017).

jasonserviss/mesenchymalSubsetAnalysis documentation built on May 17, 2019, 7:28 p.m.