Comprehensive Proteomics Metadata and Integrative Web Portals Facilitate Sharing and Integration of LINCS Multiomics Data

Abstract

The Library of Integrated Network-based Cellular Signatures (LINCS), an NIH Common Fund program, has cataloged and analyzed cellular function and molecular activity profiles in response to >80,000 perturbing agents that are potentially disruptive to cells. Because of the importance of proteins and their modifications to the response of specific cellular perturbations, four of the six LINCS centers have included significant proteomics efforts in the characterization of the resulting phenotype. This manuscript aims to describe this effort and the data harmonization and integration of the LINCS proteomics data discussed in recent LINCS papers.

Keywords: FAIRness; LINCS data portal; LINCS proteomics metadata; P100 data; metadata harmonization; piNET.

Graphical abstract

Fig. 1

Overview of multi-institutional LINCS proteomics data centers and assays. On the left, the list of participating centers in LINCS proteomics working group: BD2K-LINCS DCIC: Department of Molecular and Cellular Pharmacology, University of Miami, Miami, FL 33146; Department of Environmental and Public Health Sciences, University of Cincinnati, Cincinnati, OH 45220; Department of Pharmacological Sciences, Icahn School of Medicine at Mount Sinai, New York, NY 10029, DToxS: Department of Pharmacological Sciences, Icahn School of Medicine at Mount Sinai, New York, NY 10029, HMS LINCS Center: Harvard Medical School, Boston, MA 02115, NeuroLINCS: Cedars-Sinai Medical Center, Los-Angeles, CA 90048; Gladstone Institute of Neurological Disease and the Departments of Neurology and Physiology, University of California San Francisco, San Francisco, CA 94158; Department of Biological Engineering, MIT, Cambridge, MA 02142; Department of Neuroscience, Johns Hopkins University, Baltimore, MD 21205; Departments of Psychiatry and Human Behavior and Neurobiology and Behavior, University of California Irvine, Irvine, CA 92697, PCCSE: The Broad Institute of Harvard and MIT, Cambridge, MA 02142; Department of Genome Sciences, University of Washington, Seattle, WA 98195; Picower Institute for Learning and Memory, Department of Brain and Cognitive Sciences, Massachusetts Institute of Technology, Cambridge, MA 02139. On the right, the graphical summary of LINCS proteomics assays, cell lines, and perturbation types profiled as part of LINCS and described in this paper, as well as LINCS-related tools and resources providing platforms for proteomics data annotation and sharing: piNET (1), Panorama (2, 3), phosphoSitePlus (4, 5), The Protein Information Resource (6), and HMS LINCS (7).

Fig. 2

Graphical illustration of exploring process in LINCS Proteomics Landing Page (LPLP). A, on the top, LPLP provides case motivated search and query types, e.g. using chemical similarity to a query compound in order to find its LINCS analogs, or using the protein network neighborhood of a query gene/protein in order to identify potentially relevant genetic perturbations for which proteomics signatures are available. On the bottom, the search query is directed to different sections and other tools and resources such as LINCS Data Portal. B, example showing the exploration of Staurosporine P100 signatures in LINCS Proteomics Landing Page. In this example, (i) P100 assay is shown as a cluster of samples where pronounced patterns in the peptide levels can be (ii) zoomed in, (iii) selected and enlarged, and the average signature of such selected profiles is then sent to piNET for further analysis. (iv) A circular phospho-network representation of the P100 assay is generated using known and predicted kinase-substrate relationships, illustrating a pattern of a lower abundance for multiple phosphopeptides targeted by kinases that are in turn inhibited by staurosporine, a non-specific kinase inhibitor.

Fig. 3

Multimodal analysis of chemical perturbation signatures. A, Selecting the most robust signature from LPLP P100 signatures, identifies two CDK inhibitors. B, Submission of the phospho-peptide signature of the selected cluster to piNET identifies CDK kinase as an active kinase. C, Dinaciclib was identified as a highly concordant signature to CDK gene knockdown signatures using iLINCS. D, Dinaciclib is suggested as an active Kinase using Enrichr enrichment analysis of CDK gene knockdown signatures. E, Dinaciclib shows a stronger phospho-peptide signal compared to Dasatinib (a non-specific kinase inhibitor).