lincs: LINCS Z-score Signature Database

lincsR Documentation

LINCS Z-score Signature Database

Description

The L1000 assay, used for generating the LINCS data, measures the expression of 978 landmark genes and 80 control genes by loading amplified mRNA populations onto beads and then detecting their abundance with a fluorescent-based method. The expression of 11,350 additional genes is imputed from the landmark genes by using as training data a large collection of Affymetrix gene chips.

The LINCS data have been pre-processed by the Broad Institute to 5 different levels and are available for download from GEO. Level 1 data are the raw mean fluorescent intensity values that come directly from the Luminex scanner. Level 2 data are the expression intensities of the 978 landmark genes. They have been normalized and used to impute the expression of the additional 11,350 genes, forming Level 3 data. A robust z-scoring procedure was used to generate differential expression values from the normalized profiles (Level 4). Finally, a moderated z-scoring procedure was applied to the replicated samples of each experiment (mostly 3 replicates) to compute a weighted average signature (Level 5). For a more detailed description of the preprocessing methods used by the LINCS project, readers want to refer to the LINCS user guide.

Disregarding replicates, the LINCS data set contains 473,647 signatures with unique cell type and treatment combinations. This includes 19,811 drug-like small molecules tested on different cell lines at multiple concentrations and treatment times. In addition to compounds, several thousand genetic perturbations (gene knock-downs and over expressions) have been tested. Currently, the data stored in this package is restricted to signatures of small molecule treatments across different cells lines. However, users have the option to assemble any custom collection of the LINCS data. For consistency, only signatures at one specific concentration (10uM) and one time point (24h) have been selected for each small molecule in the default collection. These choices are similar to the conditions used in primary high-throughput compound screens of cell lines. Since the selected compound concentrations and treatment duration have not been tested by LINCS across all cell types yet, a subset of compounds had to be selected that best met the chosen treatment requirements. This left us with 8,104 compounds that were uniformly tested at the chosen concentration and treatment time, but across variable numbers of cell lines. The total number of expression signatures meeting this requirement is 45,956, while the total number of cell lines included in this data set is 30.

Details

The LINCS sub-dataset, filtered and assembled according to the above criteria, can be downloaded from Bioconductor’s ExperimentHub as HDF5 file. In this case the loaded data instance includes moderated Z-scores from DE analyses of 12,328 genes for 8,140 compound treatments across a total of 30 cell lines corresponding to 45,956 expression signatures. This data set can be used by all set-based and correlation-based GESS methods provided by the signatureSearch package.

The loaded lincs data object is filtered and generated from the original LINCS level 5 data stored at GEO: https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE92742. For documentation and code of generating the lincs databases from sources, please refer to the vignette of this package by running browseVignettes("signatureSearchData") in R.

References

LINCS source data at 5 levels from GEO: https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE92742

LINCS User Guide v2.1: https://docs.google.com/document/d/1q2gciWRhVCAAnlvF2iRLuJ7whrGP6QjpsCMq1yWz7dU/edit#

Examples

library(ExperimentHub)
# eh <- ExperimentHub()
# query(eh, c("signatureSearchData", "lincs"))
# lincs_path <- eh[['EH3226']]
# rhdf5::h5ls(lincs_path) 

yduan004/signatureSearch_data documentation built on April 7, 2023, 5:02 a.m.