LINCS L1000 measures the transcriptional response to perturbations. However, a single perturbagen is often assessed at several conditions, such as dosages, timepoints, or cell lines. A consensus signature meta-analyzes several input signatures and condenses them into a single output signature.We've computed consensus signatures for:+ 1,170 DrugBank compounds+ 4,326 gene knockdowns+ 2,413 gene overexpressions+ 38,327 L1000 perturbationsEach signature contains dysregulation z-scores for 7,467 genes (978 measured and 6,489 inferred, see genes.tsv). The consensi-{type}.tsv.bz2 files contain the perturbagen × gene matrix of z-scores. The dysreg-{type}.tsv files contain significantly dysregulated genes. The dysreg-{type}-summary.tsv files provide the counts of significantly up/down-regulated genes per perturbagen.Our methods are available on Thinklab. The project GitHub repository contains all of the datasets here besides consensi-pert_id.tsv.bz2 due to its large file size.If using these datasets, please attribute this figshare deposition and the LINCS L1000 project. Also please abide by the data release policy for the NIH LINCS Program.This is not an official LINCS L1000 repository. Users are warned that our modifications may have introduced errors or removed signal that was present the original data. We thank the L1000 team for posting their data and providing support including online office hours.

Transcriptional profiles of multiple cell and perturbation types: cells are treated with chemical perturbagens and CRISPR reagents. The expression level for 978 representative genes is measured.

The Library of Integrated Cellular Signatures (LINCS) is an NIH program which funds the generation of perturbational profiles across multiple cell and perturbation types, as well as read-outs, at a massive scale. The LINCS Center for Transcriptomics at the Broad Institute uses the L1000 high-throughput gene-expression assay to build a Connectivity Map which seeks to enable the discovery of functional connections between drugs, genes and diseases through analysis of patterns induced by common gene-expression changes. This is Level 5 data: GSE70138_Broad_LINCS_Level5_COMPZ_n118050x12328_2017-03-06.gctx.gz Series GSE70138 L1000 data is provided at five levels of the data processing pipeline: Level 1: Raw unprocessed flow cytometry data from Luminex (LXB) Level 2: Gene expression values per 1000 genes after deconvolution (GEX) Level 3: Quantile-normalized gene expression profiles of landmark genes and imputed transcripts (Q2NORM or INF) Level 4: Gene signatures computed using z-scores rela

LINCS L1000 measures the transcriptional response to perturbation. l1000.db is a SQLite database that we use to store LINCS L1000 data. Our database contains several parts of the L1000 resource that have been converted into user-friendly tables. The database also provides information not available from L1000. Namely, we computed the chemical similarities between L1000 compounds and mapped the L1000 compounds to external vocabularies.+ The cells table contains the L1000 cell lines.+ The sigs table contains the L1000 signatures.+ The perts table contains the L1000 perturbations.+ The similarity table contains chemical similarity scores for all compound pairs.+ The unichem table contains a mapping of L1000 compounds to external chemical vocabularies.This database does not contain the L1000 expression data, hence we consider it metadata. l1000.db.bz2 is bzip2 compressed to reduce file size. Decompress before use.We rely on this database when computing consensus signatures for perturbations.If using this database, please attribute this figshare deposition and the LINCS L1000 project. Also please abide by the data release policy for the NIH LINCS Program.

L1000 gene expression data collected in A549 cell lines for select drug compound perturbations. We used this dataset to compare information content in different profiling technologies.

Transcriptional profiles of cultured human breast and prostate cancer cell lines treated with small molecules: 6 cancer cell lines are treated with 6 small molecules, as a part of the LINCS Trans-Center Project. The expression level for 978 representative genes is measured.

In this fileset we provide our processed forms of the LINCS L1000 data. Starting with signature-level expression data from the L1000 project, we employ our pipeline (link below) to condense this information to consensus drug-level expression signatures.

For a detailed overview of our approach, see the appended thinklab discussion link. In brief, our approach has 3 steps. (1) Pool set of LINCS gold signatures corresponding to drug of interest. (2) Compute weights for each signature, based on their mean Spearman correlation values with all other signatures. (3) Combine Z-scores across all signatures using Stouffer's method to create a consensus signature.

These drug-level consensus signatures are provided in two drug vocabularies: LINCS perturbagens and DrugBank compounds. Rows are drugs, and columns are genes (Entrez Gene IDs). The values in both files are Z-scores, as per the LINCS L1000 standard.

Transcriptional profiles of multiple cell and perturbation types: cells are treated with chemical and genetic perturbations. The expression level for 978 representative genes is measured.

Transcriptional profiles of multiple cells treated with small molecule perturbagens: The expression level is measured for 978 representative genes in 15 cell lines treated with 241 small molecules.

modzs.gctx contains a probe (rows) × signature (columns) matrix of differential expression z-scores from the LINCS L1000 project. The file uses a compressed binary encoding that can be read using the cmap/l1ktools drivers.modzs.gctx is an input for the dhimmel/lincs repository, but is too large to be uploaded to GitHub. dhimmel/lincs extracts signatures from modzs.gctx to compute consensus signatures for perturbagens. Consensus signatures from dhimmel/lincs are available at https://doi.org/bp22 as user-friendly TSV files. modzs.gctx is uploaded for reproducibility of dhimmel/lincs and longterm preservation. Users who need raw LINCS L1000 data are encouraged to use the Broad Production Data hosted on GEO rather than the legacy modzs.gctx file.I (Daniel Himmelstein) originally retrieved modzs.gctx from the following path on the L1000 C3 Cloud (c3.lincscloud.org): /xchip/cogs/data/build/a2y13q1/modzs.gctx. However, the original hosting is no longer operating. Therefore, I am uploading this file to figshare.I am choosing a CC BY license with the requirement to attribute the LINCS L1000 Team at the Broad Institute. However, users should note that the licensing of modzs.gctx is murky. The Broad license (https://git.io/vieCF) states that data files cannot be redistributed, which conflicts the NIH LINCS Data Release Policy. See https://doi.org/bfmn for more information.

Transcriptional profiles of cultured human cancer cell lines treated with small molecules: Expression of 978 representative genes is measured in 39 cell lines perturbed by 13 thousands of small molecules.

Cell lines with the highest number of available signature profiles in the LINCS L1000 data and their corresponding annotation according to the Cell Service API.

Dataset of cellular signatures that catalogs transcriptional responses of human cells to chemical and genetic perturbation. CMap contains perturbagens, expression signatures, and small molecules from cell lines.

320 drugs are studied in SIDER dataset and 323 drugs are used in FAERS dataset. Those drugs are found in both these two datasets and the selected low quality LINCS L1000 dataset. (XLSX)

Dataset extracted from the L1000 mRNA profiling assay (LINCS from the NIH), phase 1 (http://identifiers.org/lincs.data/LDG-1227) and phase 2 (https://doi.org/10.1016/j.cell.2017.10.049) and used for the study "Linking chemicals, genes and morphological perturbations to diseases" (Cerisier et al., 2023, Toxicology and Applied Pharmacology). It contains data on 12328 genes for the 9989 compounds mentioned in the study. It contains 178925 rows; one row corresponds to one experimental condition: one compound tested on one cell type at one dose and one time. The column “hash” represents this condition: “compounds=celltype=dose=time=Inf/Sup”. The Inf/Sup information represents the sign of the z-score, If the z-score is negative and below -2.00, the symbol “Inf” is marked. If it is positive and greater than 2.00, the symbol “Sup” is marked. Z-scores between -2.00 and 2.00 are not taken into account, they do not mean any gene disturbance and they do not appear in the dataset. The "genes" column corresponds to the genes (their EntrezIDs), separated by a comma, that are perturbed by the corresponding condition in the "hash" column. The “inchikey” column is basically the inchikey of the molecules concerned by the "hash" column and the “BROAD_ID” column corresponds to the molecule ID provided by the Broad Institute.

LINCS L1000 characteristic direction signatures search engine. Software tool to find consensus signatures that match user’s input gene lists or input signatures. Underlying dataset is LINCS L1000 small molecule expression profiles generated at Broad Institute by Connectivity Map team. Differentially expressed genes of these profiles were calculated using multivariate method called Characteristic Direction.

Project to create network based understanding of biology by cataloging changes in gene expression and other cellular processes when cells are exposed to genetic and environmental stressors. Program to develop therapies that might restore pathways and networks to their normal states. Has LINCS Data Coordination and Integration Center and six Data and Signature Generation Centers: Drug Toxicity Signature Generation Center, HMS LINCS Center, LINCS Center for Transcriptomics, LINCS Proteomic Characterization Center for Signaling and Epigenetics, MEP LINCS Center, and NeuroLINCS Center.

Transcriptional profiles of U2 OS cell line treated with ~21 thousands of compounds from diversity-oriented-synthesis (DOS) library.

Supplementary Material 1

Version 1.0.0 (12/28/2022)

This is a drug repurposing dataset under MIT licence, compiled by Dr. Clémence Réda , comprising a drug-disease association matrix, and several drug-drug and disease-disease similarity matrices. It only uses transcriptomic data (i.e., gene activity/expression). The sparsity number is the percentage of nonzero values in the association matrix.

drugs | # diseases | Sparsity number | # positive associations | # negative associations | # genes

------- | ---------- | --------------- | ----------------------- | ----------------------- | ------- 871 | 144 | 0.76% | 773 | 181 | 10,811

All drugs (resp., diseases) are associated with a gene expression feature vector of length 10,811 (that is, all drugs and diseases in the feature matrices appear in the association matrix, and vice versa). However, some diseases/drugs are not necessarily involved in negative or positive associations (meaning that all pairs with those items have an association value of 0).

This dataset consists of three .CSV files:

• Drug-Disease Association Matrix

1. "ratings_mat.csv"

This matrix contains values in {-1,0,1} where -1 stands for a negative association (i.e., the drug failed for some reason to treat the considered disease: e.g., lack of accrual in the associated clinical trial, or proven toxicity), 1 for a positive association (i.e., the drug was shown to treat the disease), and 0 for unknown associated status. The columns are diseases, identified by their MedGen Concept ID, whereas rows are drugs, identified by their DrugBank IDs or PubChem CIDs.

• Drug Feature Matrix

1. "items.csv"

This matrix has drugs in its columns, identified by their DrugBank IDs or PubChem CIDs, and genes in its rows, identified by their HUGO Gene Symbol. Genewise transcriptomic variation induced by drug treatment, from the CREEDS or the LINCS L1000 databases.

• Disease Feature Matrix

1. "users.csv"

This matrix has diseases in its columns, identified by their MedGen Concept IDs, and genes in its rows, identified by their HUGO Gene Symbol. Genewise transcriptomic variation induced by the disease, from the CREEDS database.

Further information about the generation of those matrices is available by running the Jupyter notebook TRANSCRIPT_dataset-v1.0.0.ipynb on the following GitHub repository: https://github.com/RECeSS-EU-Project/drug-repurposing-datasets. For any questions, please contact the author at or the RECeSS project contributors at .