A dimensionality reduction technique particularly well suited for visualizing data. (For references, see https://lvdmaaten.github.io/tsne)

The parameters that were used for running t-SNE here are: 50 initial dimensions, perplexity of 30, and theta of 0.5. For datasets with <= 5000 samples, the standard t-SNE algorithm is used. For larger datasets, the Barnes-Hut algorithm is employed.

A dimensionality reduction technique in which the two principal components are chosen to have the largest possible variance.

To analyze relationships between perturbations, we utilize the framework of connectivity. A connectivity score between two perturbations quantifies the similarity of the cellular responses evoked by these perturbations. A score of 1 means that these two perturbations are more similar to each other than 100% of other perturbation pairs. A score of -1 means that these two perturbations are more dissimilar to each other than 100% of other perturbation pairs.

See a heatmap of connections between individual perturbagens in cell lines and all other perturbagens used for the P100 assay or the GCP assay. The tutorial describes the features of the heatmap.

Bring data, in GCT format, from your own P100 or GCP studies to query against our datasets.

Introspect means querying your dataset against itself. Make sure to "Include Introspect" if you would like to see connections within your dataset (in addition to connections between your dataset and Touchstone-P).

In computing connectivity, biological or technical replicates can be aggregated together. Please select which metadata fields should be used to recognize replicates. For example, if you wish to distinguish between different doses of the same compound, make sure to select "pert_dose" (or something similar) as one of the metadata fields by which to group replicates. The possible metadata fields by which to group replicates only appear after you have upload your GCT and selected "Yes" for "Are there replicates in your data?".


Matched mode: When running GUTC, incorporates cell-line information to match query data against matching cell types in Touchstone. Currently this includes the following 9 cell types : [A375, A549, HEPG2, HCC515, HA1E, HT29, MCF7, PC3, VCAP].
Unmatched mode (recommended): When running GUTC, does not incorporate cell-line information when querying the data against Touchstone signatures.


L-Build ("Light" Build):  All levels of L1000 data up to aggregated signatures.
Full Build:  All levels of L1000 data up to aggregated signatures, as well as all relevant additional analyses of the data (Introspect, t-SNE, PCA, etc.).

When querying Touchstone, Feature Space determines what set of genes to query against. When perturbagens are profiled on the L1000 platform, Landmark is recommended. When the queries you wish to use are not landmarks, use BING instead.

Root location within a brew folder that contains the instance matrices and the brew_group folder. Default is brew/pc

List of expected treatment doses in micromolar as a listmaker list. If provided, dose discretization is applied to the pert_dose metadata field to generate a canonicalized pert_idose field. Note this assumes that the pert_dose annotations are in micromolar.

Generates TAS plots and connectivity heatmap of preliminary callibration plates to identify the most suitable experimental conditions of specified parameters. Tool should be run on small pilot experiments, with a variety of experimental parameters such as seeding density and time point. Plots can also be decoupled by parameters such as cell id.

Column filter to sig_build_tool as a listmaker collection

The name of the build used when generating all associated files and folders (e.g. <BUILD_CODE>_metadata). For this reason, the code must be filename compatible.

When merging replicates for L1000, several versions of the merged data are made. This parameter determines which version to use when creating your build. by_rna_well is the default. by_rna_well is recommended.

All data is from the Cancer Cell Line Encyclopedia resource. Expression data was released 15-Aug-2017, copy number data is dated 27-May-2014, and mutational data is dated 15-Aug-2017.


Feature Mapping: Ensembl Ids from the source data were mapped to Entrez Gene Ids using gene annotations from NCBI (downloaded on 02-Mar-2016).
Normalization:  RNAseq RPKM values were log2 transformed using log2(max(RPKM, eps)). The data were then normalized such that the expression values were comparable across cell lines, by minimizing technical variation and equalizing their distributions (for details of the normalization, see LISS and QNORM entries in the Connectopedia glossary). Post-normalization, the expression values range between 4 and 15 log2 units, with 4 indicating that a gene is minimally or not expressed and 15 indicating the maximum readout.
Z-scores: The number of standard deviations that a gene is above or below the population mean is called its z-score. The "robust" z-score is resistant to outliers by using median instead of mean and median absolute deviation (MAD) instead of standard deviation. The reference population used to compute the median and MAD for a particular gene is all CCLE lines with data for that gene.
Z-scores Within Primary Site: Similar to z-scores, but the reference population used to compute the median and MAD is all CCLE lines from the same lineage with data for that gene.

All scores indicated are in log 2 ratios to reference, binned using the heuristics described in CNVkit.

Deletion:  score < -1.1
Loss:  -1.1 ≤ score ≤ -0.25
No change:  -0.25 < score < +0.2
Gain: +0.2 ≤ score < +0.7
Amplification: +0.7 ≤ score

Access a suite of analysis apps by clicking on the menu (or type command-K to open)

Switch between running a single query and running a batch query.

Give each query a descriptive name that will help you identify your results.

Tip: Each list can have a different number of genes; in fact, you can run a query with only one list (up OR down).

Your query will take about 5 minutes to process; check the History section in the Menu for your results!

Valid genes used in the query have HUGO symbols or Entrez IDs and are well-inferred or directly measured by L1000 (member of the BING gene set). Valid genes not used in a query are those that have a valid HUGO or Entrez identifier but are not part of the BING set. Invalid genes do not have HUGO or Entrez IDs.

Give each query a descriptive name that will help you identify your results.

Your query will take about 5 minutes to process; check the History section in the Menu for your results!

The sig_fastgutc_tool is a reimplementation of our query algorithm that enables faster query results, especially at larger batch sizes. It is the result of crowd-sourced contest. It is currently in beta mode.

Filter datasets by category to see only those of interest.

Data Icons identify published and proprietary datasets.

Click on a row to see a summary of that dataset, including cell lines and treatment conditions, assay type, and dates.

Arrange the table to display the information most important for your work, and add key datasets to favorites.

View details about the collection as a whole and about individual compounds.

View subsets of compounds based on mechanism, drug target, or known disease application.

Purity is assessed by ultra-performance liquid chromatography-mass spectrometry (UPLC-MS) of compounds after receipt from the vendor.

Status as of publication of this resource (March 2017). We will be updating this but let us know if you notice a discrepancy.

Click on a compound to see details about its structure, mechanism, targets, approval status, and vendor.

Mouse over this graphic to see the classes of proteins targeted by drugs in the hub.

This is the current count of perturbagens in the reference (touchstone) dataset.

Select data from perturbagens grouped by their MoA or role in the cell.

Choose a perturbagen type, or view them all.

Touchstone is our reference dataset, made from well-annotated perturbagens profiled in a core set of 9 cell lines.

Detailed List is unavailable for Touchstone v1.1.1.1. A new data visualization approach is in development, but to get results in a table format (similar to Detailed View), please click on Heat Map and download the dataset as a GCT file that can be viewed in Excel or similar apps. Please see here for a detailed explanation.

Articles are tagged with topics. Click on a topic tag to see all related articles.

Look it up! A quick reference guide of CMap terms and their meanings.

Email us with your questions.

Click on the heading to read all the articles in this section on a single page, or open each article separately.

Click on a heading to open a menu of articles.

Each article is tagged with key words that describe its content.

Underlined words link to their definition in the CMap glossary.

Your feedback helps us make Connectopedia more useful.

Average transcriptional impact

TAS is a metric that incorporates the signature strength (the number of significantly differentially expressed transcripts) and signature concordance (the reproducibility of those changes across biological replicates) to capture activity of a compound. The score is computed as the geometric mean of the signature strength and the 75th quantile of pairwise replicate correlations for a given signature. Prior to computing the geometric mean, the signature strength is multiplied by the square root of the number of replicates. This serves to mitigate score shrinkage with increasing replicate number and allows TAS values derived from signatures of different numbers of replicates to be compared with each other.

Signature diversity

Thick black bars signify Transcriptional Activity Scores greater than or equal to 0.5; thinner black bars denote scores less than 0.5. Absence of a bar means no data available. Colored lines (chords) signify similar connectivity scores between cell lines; red for positive connectivity scores of 80-100 (pale to intense color according to the score); blue for negative connectivity. Chords are only shown when TAS scores are > 0.5; thus absence of a chord either means that the perturbagen TAS score is very low, or that no data is available. Chords for individual cell lines can be isolated from the rest of the figure by hovering over the cell line name.

Baseline expression of this gene in each cell line is represented as a z-score (top numbers). Scores were calculated using robust z-score formula:

z-scorei = ( xi - median( X ) )/( MAD( X ) * 1.4826 ),

where:

xi is expression value of a given gene in i-th cell line

X = [ x1, x2 ... xn ] is a vector of expression values for a given gene across n cell lines

MAD( X ) is a median absolute deviation of X

1.4826 is a constant to rescale the score as if the standard deviation of X instead of MAD was used

Median and MAD expression values were calculated using RNA-Seq profiles from a total of 1022 cell lines, comprising data from the Cancer Cell Line Encyclopedia (CCLE; Barretina, et al.) and cell lines nominated by the CMap team. Plots show z-score values only for the core LINCS lines used by CMap in L1000 experiments. Light red or light blue regions indicate positive or negative outlier expression, respectively, of the gene relative to the other lines shown; z-score of a positive outlier in the corresponding cell line is in dark red and a negative outlier is in dark blue.

Summary class connectivity shows a boxplot that summarizes the connectivity of a class. Each data point, shown as a light gray dot, represents the median value of connectivity of one member to the other class members. (This corresponds to the median for each row, excluding the main diagonal, in the heatmap shown below.) The box is the distribution of those data points, where the box boundary represents the interquartile range, the vertical line within the box is the median, and the whiskers reflect the minimum and maximum values of the data (exclusive of extreme outliers, which may appear beyond the whiskers).

Connectivity between members of class is a standard heat map of the connectivity scores, summarized across cell lines, between members of the class, where dark red represents the highest positive scores and deep blue the highest negative scores. Individual scores are revealed to the left below the map by hovering over each cell of the map.

Class inter-cell line connectivity is a plot of the median (black line) and Q25-Q75 connectivity scores (blue area around black line) for each cell line as well as the summary scores across cell lines. In some cases perturbations have not been tested in every cell line; the absence of data is indicated by a “0” for that cell line. The example shown reveals that these estrogen agonists show the strongest connectivity to each other in MCF7, a human breast cancer cell line that expresses the estrogen receptor.

Profile status

Colored portion of top bar indicates the Broad assays in which this compound has been profiled.

L1000 cell/dose coverage

For compounds profiled by L1000, cell lines and dose range for which signatures are available are indicated by dark gray bars (lighter gray bar indicates no data is available for that cell line/dose combination). A bar displayed one row above the 10 uM row indicates that doses higher than 10uM were tested. The 6 rows correspond to 6 canonical doses: 20 nM, 100 nM, 500 nM, 1 uM, 2.5 uM, and 10 uM. (In some cases non-canonical doses were tested; these are rounded to the nearest canonical dose for the purpose of this display. For example, if the dose tested was 3.33uM, the 2.5uM bar is shown in dark gray here.)

Competitions
& Challenges
  Jump to:  Past ContestsAbout CMap

Past Contests

Contest 4: Mechanisms of Action (MoA) Prediction Challenge

September 4, 2020 - November 30, 2020


The goal of this contest was to develop algorithms capable of predicting a drug’s mechanism of action from its pattern of cell viability and gene expression across 100 cancer cell lines.

The training dataset consisted of readouts for over 5,000 different drugs from the  Drug Repurposing Library  that were profiled at different time points and doses. Submissions were scored for accuracy on a pre-specified metric (the logarithmic loss function, also known as cross entropy, averaged over all mechanisms).

More than 4,000 teams made code submissions for a total of 88,000 solutions evaluated with our metric. The top four competitors won cash prizes for a total of $30,000.

The top algorithms used various ensembles of a variety of different neural network models, such as Convolutional Neural Networks. The top algorithms were capable of predicting between 60 and 100 targets with higher accuracy than the benchmark (a random forest algorithm), thus outperforming our previous effort on the same data.

The challenge was hosted by Kaggle.  See the competition’s website  for more details.

Contest 3: CMap Gene Deconvolution Challenge

December 14, 2018 - January 18, 2019


This challenge has concluded.We are grateful to all of the participants who developed a variety of intriguing solutions to address the deconvolution problem. The results of the challenge are summarized in  this preprint manuscript and the contest data are available in the  clue data library.
The docker container used for converting the deconvoluted data to differential expression values can be downloaded from  docker hub.


Aid biomedical discovery by predicting the expression of two genes from a composite measurement and compete for $23,000 in prizes

A key factor in enabling the scale-up of the CMap dataset is the practice of measuring the expression of two genes using the same physical material in the L1000 assay, thus dramatically reducing the costs and increasing the throughput of data generation. In the third challenge in the CMap series, we seek to improve the speed and accuracy of ‘d-peak’, the algorithm that deconvolutes the composite expression signal into two values and associates them with the appropriate genes.

For the purpose of this challenge, the core CMap technology can be described as follows. In a single experiment, CMap makes 488 measurements. Each measurement produces an intensity histogram (a vector of integers), which characterizes expression of two distinct genes in the sample (for a total of 488 x 2 = 976 genes). In the ideal case, each histogram consists of two peaks (see Figure above), each corresponding to a single gene. The genes are mixed in 2:1 ratio, thus the areas under the peaks have 2:1 ratio, which allows us to associate each peak with the specific gene. The median position of each peak corresponds to the gene’s expression level, and that's what you need to determine in this challenge.

Contest 2: Algorithm Speed Challenge

January 26 - February 16, 2017


The goal of this context was to improve the speed of the CMap query algorithm.

Overview of the Connectivity Map

CMap enables the discovery of functional connections between drugs, genes and diseases through the generation and analysis of gene expression signatures, where each signature represents the transcriptional response of human cells to chemical or genetic perturbation. To identify connections, a researcher poses a biological question in the form of a “query” comprised of a list of genes of interest. The query algorithm then searches the CMap database to identify signatures that are most similar to the user's input. By using this algorithm and the Connectivity Map dataset, researchers have uncovered novel biological relationships and generated hypotheses for the development of new therapeutics.

The Challenge

The CMap L1000 assay quantifies the responses of 10,174 genes to an experimental perturbation. Due to recent technological improvements, it has become possible to massively scale-up data generation. As a result, the CMap matrix has grown to 476,251 signatures and this number is expected to continue to increase rapidly. The goal of the contest is to improve the speed of the query algorithm, as this has great practical importance for researchers using CMap.

Participation and Implementation

In addition to helping advance biomedical research, the winners were awarded cash prizes, with a total purse of $20,000. The first place submission showed as much as a 100x improvement over the current query algorithm. An implementation of the winning submission is now available at clue.io/query as a replacement to the current CMap Query algorithm. Improvements to overall Query result time on CLUE, which includes some additional analysis such as GUTC, range from 2-4 fold with greater improvement seen in large batch queries. Overall, the contest removed query as a bottleneck in Query app compute time and allowed for the enabling of batch queries in the Query app on CLUE - a highly requested feature by users.

Contest 1: Inference Challenge

June 28 - July 19, 2016


The goal of this contest was to maximize the accuracy of the inferred gene expression values used by the Connectivity Map, while minimizing the number of the measured gene expressions. Results of this contest expanded research horizons for computational biologists and scientists who seek to find drugs that cure diseases.

Contest Details

CMap utilizes a novel, high-throughput gene expression profiling technology to generate gene expression profiles at scale. The crux of this approach is that instead of measuring all ~20,000 genes in the human genome, CMap measures a select subset of approximately 1,000 genes and uses these “landmark” gene measurements to computationally infer a large portion of the remainder. The current algorithm is effective but imperfect, and improving the imputation methods will have an immediate impact on the quality of data and the biologically meaningful connections that can be discovered. With this in mind, we have designed our first contest to stimulate the exploration of new and improved inference methods.

Participation and Implementation

Several of the top contestants achieved a notable improvement over the current inference model. Though contestants achieved improved accuracy relative to the ground truth RNAseq dataset, this improved accuracy did not reliably translate to improvements in downstream analyses such as connectivity. One potential reason for lack of improved connectivity is that the comparison to RNAseq, using normalized profiles, may be too far removed from connectivity, which is performed using replicate-collapsed differential expression signatures Further exploration of potential applications of the improved inference is ongoing. An implementation of the contestant solutions in R are available here in the cmapR library. Click here for more details on the results, or click here to see the final leaderboard on the TopCoder site.

About CMap

The Connectivity Map (CMap) is a collection of genome-wide transcriptional expression data from cultured human cells treated with bioactive, small molecules and pattern-matching algorithms. When these elements are brought together, the results enable the discovery of functional connections between drugs, genes and diseases through the transitory feature of common gene-expression changes. For more on CMap, click here.