Q & A
Q: What is the overall goal of the program?
- LINCS is working to establish a new understanding of health and disease through an integrative approach that identifies patterns of common networks and cellular responses (called cellular signatures) across different types of tissues and cells in response to a broad range of perturbations.
- The underlying premise of the LINCS program is that disrupting any one of the many steps of a given biological process will cause related changes in the molecular and cellular characteristics, behavior, and/or function of the cell – also known as the cellular phenotype. A cellular phenotype, in turn, can be reflected by signatures derived from comparable assays of clinical states. Observing how and when a cell phenotype is altered by specific stressors can provide clues about the mechanisms involved in perturbation and, ultimately, disease.
Q: What are cellular signatures?
- A cellular signature of a perturbagen response is the set of reduced dimensionality descriptors of the underlying data that provide insight into mechanism and serve as predictors. Therefore, meaningful signatures are dependent on the assay, and on how diverse assays are integrated together, either into predictive patterns or signaling networks that could lead to mechanistic interpretations.
- To develop meaningful signatures, data from diverse assays must be scaled and normalized. Integration, normalization and scaling of heterogeneous, multi-parameter dose-response data is a non-trivial task involving conceptual and practical hurdles.
Q: What assays are being run?
- The program has two data generation centers, one at the Broad Institute and one at Harvard Medical School (HMS). The Broad assays monitor gene expression using a novel ligation-mediated amplification (LMA) with a Luminex-bead based detection system to allow the quantitation of 1000 mRNA transcripts per well. Details of the assay can be found here. The HMS center monitors cell responses using multiple biochemical and cell biological assays. They range from direct assays of drug-kinase interaction in cell extracts, to multiplex biochemical assays of cell signaling proteins, to imaging assays, to assays of transcriptional response (in collaboration with the Broad LINCS Center) and cell viability. More details can be found here.
Q: What perturbations are being used?
- The program uses small molecules, ligands, and gene over-expression and knockdown (RNAi) as perturbations.
- Each data generation center has its own set of perturbations and they are determined primarily by the assay technology being used to determine response.
- The Broad profiles: 4,000 small-molecules, 3,000 human genes perturbed using lentivirally-delivered shRNAs, and 3,000 genes profiled for the effect of over-expression generating a ‘full’ data matrix. See lincscloud.org for more details,
- HMS currently focuses on two types of perturbagens, small molecule protein kinase inhibitors and naturally occurring growth factors, cytokines and chemokines in various combinations. These assays do not generate a ‘full’ data matrix and HMS uses an adaptive experimental design. See here, and here for more details.
Q: Is there a timeline available for release of data?
- The new bolus of data available every quarter will be summarized here.
- In addition to the data, LINCS is developing metadata standards for perturbations, assays, and protocols including minimum information standards. These are available here. Updates will be made every quarter as well.
- Metadata annotations will be available along with data release. We are working diligently to include sufficient metadata with each release of data, though it inevitably is a slower process.
Q: So where are the signatures?
- This past year work at each of the data generation centers has been focused on data generation, standardization, normalization etc.
- This focus is shifting to building signatures and making them available. Each of the data generation centers is currently working on generating signatures from their own experiments and these will be available soon using enabling queries accessed via this website.
- The signatures will be available via multiple methods: (a) queries that display signatures identified by the LINCS groups; (b) algorithms that a user can run to generate or build their own signatures; (c) availability of data and algorithms packaged together available for download that can re-generate LINCS published signatures; (d) statistically significant mechanistic networks that along with algorithms; (e) LINCS data signatures available via non-LINCS tools like the Cancer Genome Browser; (f) via the tools and user-interfaces being developed by the LINCS computational U01 groups.
Q: Is there a timeline available for querying the signatures?
- We have not developed a specific timeline for this yet, but the first set of tools will be available by Fall 2012.
Q: What are your future plans?
- In the coming months you will see multiple user interfaces to query LINCS data and signatures, and publications demonstrating the utility of the LINCS approach.
- We are also focusing on generating data in primary cells and iPS cells and data pertaining to these will be available.
- Our computational U01 and technology U01 centers are working to enable multiple resources to the community. See the individual sites for details.
Q: What data integration challenges are you taking on?
- Data is being collected via a joint project between the two data generation centers that will explore the relationship between immediate early cell signaling events and transcription. This will constitute the largest such public dataset, generated in a coherent manner, available for download and querying.
- One of the biggest challenges we have is to build standards for metadata for all of LINCS perturbations, assays and experiments. We want to annotate these consistently, and to make them available consistently so that outside groups can perform their own analysis and build better methods to extract signatures from the LINCS data. This is a continuing challenge given the specific nature of some of the LINCS assays and does not have easy solutions.
Q: Any opportunities to collaborate?
- There was a formal call for nominations from the community regarding additional cell lines
to be profiled at the Broad. We got an enthusiastic response. Additional cell lines in assay
development can be found here.
- We welcome collaborations to improve algorithms that build signatures from the LINCS data.
- We are publishing our metadata standards (here) and invite community input on metadata standards and data formats. Such standards (and data availability using these) would influence how the community might use LINCS data.
We are currently working on developing more formal collaboration mechanisms.
Q: What projects are being undertaken at the LINCS U01 centers?
- The technology U01 groups are working on a range of projects:
- At Arizona State we are working on building a high-throughput live-cell microarray to measure metabolites (details here )
- At Broad Institute we are working on building a mass-spectrometry based method to perform quantitative global phosphoproteomic profiling (details here)
- At Columbia we are working on building methods to generate molecular profile data necessary to transform combination therapy from a trial and error exercise to a rational discipline (details here)
- At Yale we are working on high-content and high-throughput analysis for > 20-plexed single cell phosphoproteomics analysis (details here).
- The computational U01 groups are working on a broad range of problems. As a group they, in collaboration with the data generation centers, act as beta testing sites for data tools release prior to public release. They along with the U54 centers work jointly on the metadata standards being developed by LINCS. All four centers are working on developing novel algorithms for integrative data analysis, specifically,
- At Cincinnati we work on integration of the same type readouts into meta-signatures and integration of various types of readouts and meta-signatures into regulatory network activity signatures (details here)
- At Columbia we work on building algorithms that use LINCS in vitro signatures to predict in vivo compound-related properties like mechanism of action, activity, sensitivity, synergy, etc. (details here)
- At Wake Forest we are working on building a software platform that would perform a range of tasks that are integral to LINCS data analysis including processing cellular images, modeling phosphoprotein cellular signaling pathways and integrating toolkits for mapping genomics and proteomics to cellular phenotypes (details here)
- At Miami we work on developing/using semantic-web technologies and domain level ontologies to make LINCS data coherent so that they are easily interpretable, and actionable for scientists of different backgrounds and with different objectives (details here).