Gene Signature Lab: Exploring integrative LINCS (ilincs) Data and Signatures Analysis Portal & Other LINCS Resources Jarek Meller, PhD BD2K-LINCS Data Coordination and Integration Center University of Cincinnati Gene Signature Lab, Comp. Genomics Course, IGB 607
Outline A couple of quick reminders: CMAP & LINCS Interacting with Big Omics Data using ilincs (Medvedovic et al.) Part I: deriving and interpreting genomic signatures P53 ER Part II: searching for drug targets and drugs Exploring other LINCS-related tools (Enrichr, L1000CDS2, Ma ayan et al.) Gene Signature Lab, Comp. Genomics Course, IGB 607
LINCS: Extending Connectivity Map Negative correlation with disease transcriptional signature Potential of the drug to reverse the disease process J Lamb et al. Science 2006;313:1929-1935
cell types LINCS Cube Cancer cell lines ips cells Primary cells Transcriptomic (L1000, RNA-seq) Proteomic Phosphoproteomic Morphoplogical Proliferation, apoptosis, Perturbations Chemical perturbagens (~30,000 x doses) Genetic perturbations (~30,000 x shrnas) Microenvironment perturbations Disease http://lincsproject.org
Towards Using CMAP/LINCS as Resources for Personalized Precision Medicine NOTE that small molecules with negatively correlating signatures with respect to an individual tumor signature (characterized by some mutations and some up- and down-regulated genes) could potentially be used to identify drugs to treat that particular tumor! This can be viewed as reversing the signature of the tumor This and other applications can be greatly facilitated by highly integrative and intuitive tools that enable seamless interaction with Big Omics Data, such as LINCS ilincs 5
ilincs: Linking Datasets and Signatures with Online Analysis What are my genes/proteins doing in other datasets? Constructing and analyzing signatures from transcriptomics and proteomics datasets Analyzing and mining perturbation and disease signatures ilincs.org, Mario Medvedovic et al., University of Cincinnati
ilincs Team ilincs.org, Mario Medvedovic et al., University of Cincinnati
ilincs Demo I: p53 Signature in Breast Tumors Gene Signature Lab, Comp. Genomics Course, IGB 607
Getting started Go to http://www.ilincs.org/ilincs/ Select Datasets workflow by either clicking on Datasets in the top bar or data sets icon below icon Select All Data sets and TCGA (click on Choose button to the right); select the 3 rd data set from the top (919 BRCAs) 9
Exploratory analysis Explore Heatmap Download Data ilincs.org, Mario Medvedovic et al., University of Cincinnati
Note that NAs can be effectively classified ilincs.org, Mario Medvedovic et al., University of Cincinnati
Let us generate p53 signature ilincs.org, Mario Medvedovic et al., University of Cincinnati
Gene Signature Lab, Comp. Genomics Course, IGB 607
P53 signature can be used to reclassify wt and mutants JM - http://folding.chmcc.org 14
Work around to generate the correct heatmap: Use the signature to re-analyze the same data set.
Work around to generate the correct heatmap
Work around to generate the correct heatmap
Avi Ma yan et al., Mount Sinai School of Medicine
Gene Signature Lab, Comp. Genomics Course, IGB 607
Big p53 signature 21
Dataset Analysis Workflow Enrichment analysis via Enricher Pathway analysis vis SPIA algorithm LINCS RNA-seq dataset Differential gene expression signature Small molecule CD signatures L1000CDS2 LINCS RPPA dataset TCGA RNA-seq BC dataset Connected TF binding and L1000 KD signatures ilincs.org, Mario Medvedovic et al., University of Cincinnati
ilincs Datasets 3,600 Datasets TCGA Transcriptomics ENCODE TF Binding Data P100 + GCP Proteomics ilincs.org, Mario Medvedovic et al., University of Cincinnati
Signatures Workflow Finding signatures Analyze Connected Signatures ilincs.org, Mario Medvedovic et al., University of Cincinnati
ilincs Signatures ilincs.org, Mario Medvedovic et al., University of Cincinnati
Genes Workflow Finding genes Dataset workflow Signatures workflow ilincs.org, Mario Medvedovic et al., University of Cincinnati
ilincs Demo II: ER Signature in Cell Lines vs. Breast Tumors Go to http://www.ilincs.org/ilincs/ Select Datasets workflow by either clicking on Datasets in the top bar or data sets icon below icon Select LINCS Data sets and select the last data set Oregon Health Sciences 54 mrna-seq samples from cell lines (click on Analyze button to the right) Click on Generate a Signature Select Grouping variable as ER Define groups as + and - (ER positive and ER negative cell lines) Click Create signature Select Use differentially expressed genes to analyze another set (work around) and choose the same Oregon Health Sciences data set and select Statistical analysis of genes and select ER again as the grouping variable, open heatmap Do the same, but this time find the TCGA BRCA data set and generate heatmap Gene Signature Lab, Comp. Genomics Course, IGB 607
Cell lines cluster largely by ER status; unassigned cell lines can be predicted to have either negative or positive ER status. Note that genes were selected to make that happen this is not a truly unsupervised approach. ilincs.org, Mario Medvedovic et al., University of Cincinnati
ilincs.org, Mario Medvedovic et al., University of Cincinnati
Going back to the page with ER signature: Step-by-step instructions one more time Go to http://www.ilincs.org/ilincs/ Select Datasets workflow by either clicking on Datasets in the top bar or data sets icon below icon Select LINCS Data sets and select the last data set Oregon Health Sciences 54 mrna-seq samples from cell lines (click on Analyze button to the right) Click on Generate a Signature Select Grouping variable as ER Define groups as + and - (ER positive and ER negative cell lines) Click Create signature Click Enrichr to perform enrichment analysis
Going back to the page with ER signature
Gene Signature Lab, Comp. Genomics Course, IGB 607
Avi Ma yan et al., Mount Sinai School of Medicine
ilincs Demo III: Reversing ER Signature Gene Signature Lab, Comp. Genomics Course, IGB 607
1 3 2
Searching by Gene Knockdown Signatures
Group Analysis of Raloxifen Signatures
Concordant vs. Discordant Signatures
Searching for Novel ER(-pathway) Inhibitors (concordance>0.4)
Caveats: Potentially Small Overlap with L1000 Gene Set for User Defined Signatures
Caveats: Current Sparse Cube Transcriptomics HMS LINCS Proteomics Microenvironment DTox http://lincsproject.org/ NeuroLINCS
Take home messages: i) Potential gold mine for hypothesis generation and mechanistic insights ii) Use with utmost caution, do not over-interpret, validate iii) Please be somewhat patient with the tools they keep getting better For the rest f the lab, try to reproduce as much as possible one more time. Gene Signature Lab, Comp. Genomics Course, IGB 607