The overall goal of the DTSGC is to use genomic and proteomic high-throughput measurements coupled with mid-throughput experimental measurement of protein states as the basis for computational analysis that integrates network analyses with structural constraints and dynamical models in multiple cell types to identify signatures that predict toxicity induced by individual drugs and mitigation of this toxicity by dru combinations. To anchor the signatures in observable human disease and therapeutics, we will leverage the strategy employed in our recent study, in which we searched the FDA-Adverse Event Reporting System Database (FAERS) and found nearly thousands of drug combinations used in humans where a second drug mitigates serious toxicity associated with first drug. We hypothesize that we can use these observations to improve our capability to predict toxicity of drugs and mitigation by drug pairs. The Center has three major goals: 1) experimentally obtain expression patterns of mRNA, proteins and protein states (e.g. phosphorylation) for around 250 perturbagens: 120 two-drug combinations identified in the FAERS whereby the second drug mitigates serious toxicities induced by the first drug and 130 individual drugs that have been shown in FAERS to cause one of three serious toxicities-cardiotoxicity; hepatic toxicity or peripheral neuropathy. We will use primary or established human cell lines and cell types directly differentiated from human induced pluripotent cells (hIPSC). For each drug combination and the two constituent drugs we will obtain mRNA, proteomic data, and dynamic protein state from least 18 cell lines. 2) We will utilize the experimental data for multi-tier analyses that combines statistical and network models using the human interactome and Gene Ontology with structural model based filtering and dynamical multi-compartment ODE models to obtain sets of relational signatures for each drug combination. For this we will combine the perturbagen induced changes in mRNA levels and protein levels to develop networks that will be constrained by structural modeling to identify new off-targets and dynamical models using the protein state data. The network models provide toxicity mitigation mechanism hypotheses in the form of a directed sign-specified graph which will be quantitatively weighted by global sensitivity analysis of the dynamical models. We propose to obtain non-weighted and weighted signatures for both drug combinations and individual drugs at the rate of around 4000 signatures per year. 3) We will develop computational and visualization tools for sharing the raw and processed data with the LINCS Data Coordinating Center and the larger community. We will i) develop web-based tools for data visualization and de novo analysis for all types of researchers ii) run web-based courses using Coursera for data utilization and development of signature-based research projects iii) conduct 4-6 personalized workshops to enable academic researchers to utilize our signatures to develop research projects that can compete for individual research grant funding.
Showing the most recent 10 out of 16 publications
