GDPx2

GDPx2 Functional Genomics Dataset: DRUG-seq + Chemical Perturbation in 4 Primary Cell Types. This dataset contains the supplementary data for the bioRxiv preprint "Mapping the Transcriptional Landscape of Drug Responses in Primary Human Cells Using High-Throughput DRUG-seq".

Experimental Design

Experiments were performed in a 384-well format assay. Four primary human cell lines were used: aortic smooth muscle cells, skeletal myoblasts, dermal fibroblasts, and epithelial melanocytes. 85 compounds were tested at 6 concentrations (3000, 900, 300, 95, 28.5, and 9.5 nM) using four replicates for each concentration. DMSO was used as an inert control added to different volumes (0.625%, 0.1875%, 0.0625%) to match the final DMSO concentrations found in compound-treated wells. Dexamethasone, Trichostatin A, and Brefeldin A were used as transcriptional controls. This generated a total of eight 384-well plates per cell line that were harvested for downstream transcriptomic analysis after 24 hours of exposure to the compounds or DMSO controls.

Metadata - metadata.csv

File with sample-level annotations and sample-level quality control metrics. The metadata table has the following fields:

Column	Type	Description
sample_id	int	Unique ID of the sample (well) prepared for sequencing
container_id	int	Unique ID of the perturbation or sequencing library plate. The same plate organization was preserved from perturbing cells with compounds to cell lysis and library preparation.
column_id	int	Column position in the plate
row_id	int	Row position in the plate
analysis_id	int	Unique ID of the primary analysis presented for this sample. This key is used for data upload and can be ignored.
is_edge	boolean	Whether a sample is in a well touching the border of the sequencing library plate
compound	string	Name of compound used to perturb cells
compound_concentration	float	Compound concentration
compound_concentration_unit	string	Compound concentration unit of measure
cell_line	string	Cell line name
timepoint	int	Time between perturbation and lysis in hours
condition	string	Combination of cell line, compound, concentration, and timepoint, omitting invariate columns within an experiment, to help group replicates.
percent_volume_dmso	float	Concentration of DMSO in perturbation sample, in percent volume. Compounds are dispensed dissolved in DMSO.
sample_type	string	Kind of sample, either "library" or the name of a control
is_neg_control	boolean	Whether a sample is considered a negative control
is_pos_control	boolean	Whether a sample is considered a positive control
seeded_cell_count	int	Count of cells seeded prior to perturbation
total_sequenced_reads	int	Total reads attributed to this sample
total_umi_count	int	Total Unique Molecule Identifier (UMI) count
sequencing_saturation	float	Sequencing saturation, calculated as 1 - (total_umi_count/total_sequenced_reads)
ngenes3	int	Count of genes with at least three UMIs
n_mapped	int	Count of UMIs mapped to annotated genes
percent_mapped	float	Percent of UMIs mapped to the reference, calculated by dividing n_mapped by total_umi_count
percent_rrna_removed	float	Percent of input reads attributed to rRNA and removed
percent_mitochondrial	float	Percent of UMIs attributed to mitochondria RNA
unassigned_multimapping	int	Number of reads that could not be uniquely assigned to a single gene due to mapping to multiple genes
unassigned_nofeatures	int	Number of reads that could not be assigned to any gene
percent_duplicated	float	Percentage of reads that are UMI duplicates

Compound Library - compound_library.csv

Additional information about the compounds tested.

Column	Type	Description
Drug	string	Name of the compound
Type	string	"positive control", "negative control", or "library"
Category	string	Functional category of the compound
Function	string	Function of the compound

Gene Counts - gene_counts.parquet

Gene-level UMI counts for each sample were obtained following rRNA removal (bbduk), STAR alignment (STAR) against hg38, deduplication using umi-tools, and counting with featureCounts. Columns headers represent sequenced sample IDs (sequenced_id in metadata table) and rows are genes. See methods section 4.3 in the pre-print for more details.

Differential Expression - differential_expression.parquet

Differentially expressed genes were identified using DESeq2, comparing each treatment (compound at a specific concentration) to DMSO controls with matched DMSO concentration on the same plate. See methods section 4.4 in the pre-print for more details.

Column	Type	Description
cell_line	string	Cell line or cell type
compound	string	Compound name
concentration	numeric	Compound concentration in nM
gene	string	Gene symbol
baseMean	numeric	baseMean value computed by DESeq2, corresponding to the average of normalized counts across samples treated with a particular concentration of compound and the matched DMSO control samples (see Methods)
log2FoldChange	numeric	log2FoldChange value computed by DESeq2, corresponding to the base-2 logarithm of the normalized count fold change between samples treated with a particular concentration of compound and matched DMSO control samples, after applying shrinkage (see Methods)
lfcSE	numeric	lfcSE value computed by DESeq2, corresponding to the standard error of the log2FoldChange estimate (see Methods)
pvalue	numeric	pvalue computed by DESeq2 representing the statistical significance of the observed differential expression (see Methods)

Cluster Memberships - cluster_memberships.csv

Cluster Memberships- Identify of compounds belonging to clusters identified by Canonical Correlation Analysis. See methods section 4.6 in the pre-print for more details.

Column	Type	Description
cell_line	string	Cell line or cell type
compound	string	Compound name
concentration	numeric	Compound concentration in nM
sample_id	numeric	Sample ID
cluster	numeric	Cluster number (0-based)
majority_cluster	numeric	Majority cluster for all samples with the same combination of cell_line, compound, and concentration. In case of a tie, the first cluster in numerical order is chosen.
fraction_in_majority_cluster	numeric	Fraction of samples sharing the same combination of cell_line, compound, and concentration belonging to the majority_cluster

Dose Response - dose_response.parquet

Genes showing dose-dependent gene expression changes identified based on logistic-regression. See methods section 4.7 in the pre-print for more details.

Column	Type	Description
cell_line	string	Cell line or cell type
compound	string	Compound name
gene	string	Gene symbol
log2fc_min	numeric	Minimum log2FoldChange calculated by DESeq2 across all compound concentrations (see Methods)
log2fc_max	numeric	Maximum log2FoldChange calculated by DESeq2 across all compound concentrations (see Methods)
fit_alpha	numeric	Fitted value computed by drda for the alpha parameter of the 4-parameter logistic model (see Methods)
fit_delta	numeric	Fitted value computed by drda for the delta parameter of the 4-parameter logistic model computed by drda (see Methods)
fit_eta	numeric	Fitted value computed by drda for the eta parameter of the 4-parameter logistic model (see Methods)
fit_phi	numeric	Fitted value computed by drda for the phi parameter of the 4-parameter logistic model (see Methods)
fit_anova_model1_aic	numeric	Akaike Information Criterion computed by drda for the 1-parameter (i.e., horizontal line) model (see Methods)
fit_anova_model2_aic	numeric	Akaike Information Criterion computed by drda for the 4-parameter logistic model (see Methods)
fit_anova_model2_vs_model1_pval	numeric	p-value of the Likelihood Ratio Test computed by drda comparing the 4-parameter logistic model to the 1-parameter model (see Methods)
fit_anova_model2_vs_model1_pval_adj	numeric	p-value of the Likelihood Ratio Test computed by drda comparing the 4-parameter logistic model to the 1-parameter model, adjusted for multiple testing by cell line and compound using the Bonferroni-Hochberg procedure (see Methods)
pass_complete_series	boolean	Whether all compound concentrations have a non-NA shrunken log2FoldChange computed by DESeq2 (see Methods)
pass_robust_log2fc	boolean	Whether at least one of DESeq2-computed log2FoldChange values is greater than 1 for genes with positive slope or lower than -1 for genes with negative slope, with the slope sign derived from the fitted delta parameter of the logistic function (see Methods)
pass_aic	boolean	Whether the 4-parameter logistic model provides a better fit than a horizontal line according to the Akaike Information Criterion (see Methods)
pass_lrt	boolean	Whether the p-value from the Likelihood Ratio Test comparing the 4-parameter to the 1-parameter model, adjusted for multiple testing, is less than 0.05 (see Methods)
is_dose_dependent	boolean	Whether all four criteria above are met (see Methods)

GSEA - gsea_hallmark_genes.csv & gsea_hdac_inhibitors.csv

Gene set enrichment analysis results for each treatment identified by ClusterProfiler. See methods section 4.6 in the pre-print for more details.

Column	Type	Description
cell_line	string	Cell line or cell type
gene_set_id	string	Gene set ID
compound	string	Compound name
concentration	numeric	Compound concentration in nM
majority_cluster	numeric	Majority cluster for all samples with the same combination of cell_line, compound, and concentration. In case of a tie, the first cluster in numerical order is chosen.
fraction_in_majority_cluster	numeric	Fraction of samples sharing the same combination of cell_line, compound, and concentrationbelonging to the majority_cluster
setSize	numeric	setSize computed by ClusterProfiler's GSEA function, corresponding to the number of genes in the gene set (see Methods)
enrichmentScore	numeric	enrichmentScore (ES) computed by ClusterProfiler's GSEA function, reflecting the gene set overrepresentation at the top (positive enrichment scores) or bottom (negative enrichment scores) of the list of genes in DRUG-seq data, ranked by decreasing shrunken log2FoldChange (see Methods)
NES	numeric	Normalized Enrichment Score (NES) computed by ClusterProfiler's GSEA function, accounting for differences in gene set size and in correlations between gene sets and DRUG-seq data (see Methods)
pvalue	numeric	pvalue computed by ClusterProfiler's GSEA function, representing the statistical significance of the enrichment score (see Methods)
p.adjust	numeric	p.adjust computed by ClusterProfiler's GSEA function, representing the statistical significance of the enrichment score adjusted for multiple testing across gene sets (see Methods)
qvalue	numeric	qvalue computed by ClusterProfiler's GSEA function (see Methods), estimating the false discovery rate for the normalized enrichment score (see Methods)
rank	numeric	rank computed by ClusterProfiler's GSEA function, corresponding to the position in the ranked list of genes where the maximum enrichment score occurs (see Methods)
leading_edge	string	Leading edge metrics computed by ClusterProfiler's GSEA function. "Tags" is the percentage of gene hits before the enrichment score peak for positive ES, or after the peak, for negative ES. "List" is the percentage of genes in the ranked gene list before the enrichment score peak for positive ES, or after the peak for negative ES. Signal is the enrichment signal strength, combining the "tags" and "list" statistics (see Methods).
core_enrichment	string	core_enrichment computed by ClusterProfiler's GSEA function, corresponding to the list of genes found in the leading edge and, therefore, contributing the most to the enrichment signal (see Methods)