inst/extdata/mRNA/README.md

In VanAndelInstitute/bifurcatoR: Hunt for bumps in high-dimensional (or low-dimensional!) data

TARGET pAML data extract

This folder contains an extract of a dozen genes tabulated via featureCounts against an ENSEMBL transcriptome and then collapsed by symbol. The logcounts assay is computed against the total counts of all the symbols before subsetting.

Files and descriptions

README.md this file

• ../../data/dxsmall.rds a 12x583 SummarizedExperiment: data(dxsmall)
• dxsmall.colData.csv the column data (subject covariates) for dxsmall
• dxsmall.rowData.csv the row data (gene symbol covariates) for dxsmall
• dxsmall.readcounts.csv read count for each gene in each subject specimen
• dxsmall.lognormcounts.csv log2(1 + (read count for gene/total read count))
• dxsmall.reassemble.R reassemble the dxsmall object from above CSV files
• dxsmall.plots.R template for plotting distributions

Sanity checking

If one runs dxsmall.reassemble.R and also reads in dxsmall from /data,

data(dxsmall)
dxsmall0 <- dxsmall
exists(dxsmall0)
rm(dxsmall)
source("dxsmall.reassemble.R")
exists(dxsmall)

then

identical(dxsmall, dxsmall0)

should return TRUE.

Columns (subjects)

This data is, more or less, a collection of chromatin modifier fusion leukemia. The subjects in the columns of the assays are pediatric acute myeloid leukemia patients with an MLL (KMT2A) fusion or NUP98 fusion. These two genes are the most promiscuous among fusion gene partners, and also some of the most lethal. Within defined subgroups (KMT2A-X, NUP98-NSD1, NUP98-KDM5A) of fusions, there remains substantial immunological, treatment response, and salvage heterogeneity that is not explained by genetics, cytogenetics, age group, sex, race, or FAB (morphological disease subtype). This heterogeneity is prognostic and predicts whether a child is likely to survive their disease with existing therapies.

In our experience, it makes sense to treat the two NUP98 fusion classes as separate types of disease, and to treat the majority of KMT2A fusions as if they are one type of disease. The reason for this is relatively simple: KMT2A fusions have a coherent underlying transcriptional signature (HOXA9, MEIS1, ...) which is sufficient to recapitulate disease in the absence of the fusion. NUP98 fusions do not have such a signature and the mechanism by which each generates leukemia is not yet clear. Evidence suggests that aberrations on chromosome 13 may explain some of the bimodality seen in KDM5A fusion leukemia, with stem-like abnormal-13q cases having substantially better responses to therapy. The extent of this relationship, and its biological mechanism, is an open research topic.

To a first approximation, then, we have two groups (MLL and KDM5A fusions) with substantial unexplained bimodality in expression, and one (NUP98-NSD1) without.

Rows (genes)

The genes chosen for this dataset include:

• five with bimodal correlated mRNA expression
• one housekeeping gene (GAPDH), two rarely expressed genes (CFTR, NCAM1/CD56)
• one highly expressed gene (IL3RA/CD123),
• three (KDM5D, PRAME, and C3orf80) with uninformative variation across subjects

MECOM and PRDM16 are particularly noteworthy for their bimodality.

Notes

• See dxsmall.plots.R for some skeletal tire-kicking and plot-making code.

• help("dxsmall") for some ideas about exploring the data with iSEE.

VanAndelInstitute/bifurcatoR documentation built on Oct. 13, 2024, 6:29 p.m.