DOSE-L1000 Visualizer

DOSE-L1000-Viz

by Junmin Wang

The LINCS L1000 project has collected gene expression profiles for thousands of compounds across a wide array of concentrations, cell lines, and time points. However, conventional analysis methods often fall short in capturing the information encapsulated within the L1000 dose-response data.

To faciliate the analysis of L1000 dose-response data, we created the DOSE-L1000 database. DOSE-L1000 was created by fitting generalized additive models (GAMs) and robust linear models (RLMs) to quantile-normalized log2-transformed gene expression data in L1000, followed by differential expression analysis and potency/efficacy calculations. Over 140 million models were fitted to a vast array of 33395 compounds, 82 cell lines, and 978 genes. Details can be found in the following publication:

  • Wang J, Novick S. DOSE-L1000: unveiling the intricate landscape of compound-induced transcriptional changes. Bioinformatics 39 (11): btad683.

Here we introduce DOSE-L1000-Viz, a shiny app that provides rich visualization capabilities for examining the DOSE-L1000 database.

How to Use Dose-L1000-Viz:

  • Dose Response Curves is a module that plots the dose reponse data and GAM-fitted curves for selected compound-gene pairs. To begin, please select a compound, followed by a gene, cell line, and time. The text following the colon symbol in the compound name is the BROAD ID. Hit the 'Add Condition' button to save it (you may select up to six conditions). Then, click 'Generate Plot' to visualize the data. The black dots are the experimental data, while the light blue curve is the model fit. Values inside the parentheses represent the 95% confidence intervals of potency and efficacy.

    - Example : To understand how bazedoxifene affects the dose response of ER targets such as BIRC5, add and plot the conditions of 'bazedoxifene : BRD-K90195324', 'breast : MCF7', '24h', and 'REP.A026' coupled with 'BIRC5'. This shows how ER targets like BIRC5 respond to bazedoxifene.

  • Efficacy vs Potency is a module that illustrates the estimates (\(\mu\)) and standard errors (\(SE\)) of efficacy (i.e., % DMSO) and potency (i.e., IC50 or EC50 depending on the direction of the change) of selected compound-gene pairs. First, select a field for comparison (either compound, gene, or cell line) to allow multiple inputs. This will be the variable you want to compare across. Next, choose the remaining fields to hold constant (control variables). Afterwards, click the 'Add Condition' button to save the selected condition (a limit of six conditions applies). Finally, click 'Generate Plot' to visualize the data. The error bars represent the 95% confidence intervals of potency and efficacy: \(\exp(\log(\mu) \pm z_{1 - \frac{0.05}{2}} \cdot \log(SE))\). A compound is considered to have significant effects on a gene if the 95% confidence interval of its efficacy does not contain the dashed line, i.e., 100% DMSO or equivalently, no change.

    - Example 1 : To understand the potency and efficacy of bazedoxifene acting on CTSD, MYC, CCND1, and BIRC5, choose 'Gene' to include multiple values. Then add and plot the conditions of 'bazedoxifene : BRD-K90195324 (REP.A026)', 'breast : MCF7', and '24h' coupled with 'CTSD', 'MYC', 'CCND1', and 'BIRC5'. This shows the quantitative differences in signal propagation to downstream targets following ER inhibition.

    - Example 2 : To understand the potency and efficacy of bazedoxifene acting on CTSD in MCF7, HA1E, HT29, and A375 cell lines, choose 'Cell Line' to include multiple values. Then add and plot the conditions of 'bazedoxifene : BRD-K90195324 (REP.A026)', 'CTSD', and '24h' coupled with 'breast : MCF7', 'kidney : HA1E', 'large intestine : HT29', and 'skin : A375'. This shows the cell line-specificity of bazedoxifene.

    - Example 3 : To understand the potency and efficacy of toremifene, clomifene, and bazedoxifene acting on CTSD in MCF7 cells, choose 'Compound' to include multiple values. Then add and plot the conditions of 'CTSD', 'breast : MCF7', and '24h' coupled with 'toremifene : BRD-K51350053 (REP.A007)', 'clomifene : BRD-K29950728 (REP.A009)', and 'bazedoxifene : BRD-K90195324 (REP.A026)'. This highlights the quantitative differences among the ER blockers.

  • Target View is a module that generates volcano plots for any target in DOSE-L1000. To start, please select a gene, followed by a cell line and time. To continue, click the 'Add Condition' button to save the selected condition (up to six conditions are allowed). Once done, click 'Generate Plot' to display the volcano plots. The dashed line represents a p-value threshold of 0.05.

    - Example : To identify PAX8-inhibiting compounds in HA1E kidney cells at 6h, add and plot the condition of 'PAX8', 'kidney : HA1E', and '6h'. This highlights the top hits capable of modulating PAX8 at an early time point.

  • Compound View is a module that generates volcano plots for any compound in DOSE-L1000. To start, please select a compound, followed by a cell line, time, and dose. To continue, click the 'Add Condition' button to save the selected condition (a maximum of six conditions can be selected). Once you have made your selections, click 'Generate Plot' to create the volcano plots. The dashed lines represent an adjusted p-value threshold of 0.05 and log2 fold change thresholds of +/-1.

    - Example : To understand how clomifene affects the transcriptome of MCF7 breast cancer cells in a dose-dependent manner, add and plot the conditions of 'clomifene : BRD-K29950728', 'breast : MCF7', '24h', and 'REP.A009' at '0.041uM', '0.12uM', '0.37uM', '1.1uM', '3.3uM', and '10uM'.

  • Signature Search is a module that identifies perturbation conditions with similar transcriptional signatures. It compares user-provided input to all 82741 reference signatures in the DOSE-L1000 database using Fisher's exact test. To start, please provide upregulated and downregulated gene sets in the input boxes (one gene per line). Click 'Validate' to check that your genes are present in the L1000 landmark gene space. Once ready, click 'Submit' to run the search.

    - Click Example to load a sample signature of pravastatin in HA1E kidney cells, 10uM, 24 hrs post-treatment.

  • Users can download the backend database as individual RDS files through the 'Download Data' tab. GSE92742 and GSE70138 are two phases of the LINCS L1000 project.

Important Notes:

  • Once the 'Generate Plot' button is clicked, a 'Download Data' button will appear. All data underlying the plots can then be downloaded as tab-delimited text files.
  • In Compound View, p-values are adjusted within each compound. However, in Target View, no adjustment is applied. Therefore, please use unadjusted p-values only for ranking compounds, not for assessing statistical significance.
  • In Target View, only the top 1000 perturbation conditions (compound and dose) are displayed in the volcano plot. The rest of the conditions can be found in the downloadable data.
  • Please contact Junmin at jmwang.bio@gmail.com should you have any questions or comments.