Abstract
Genome-Wide Association Studies (GWAS), whole genome sequencing, and high-throughput techniques have
generated vast amounts of diverse omics and phenotypic data. However, these sets of data have not yet been
fully explored to improve the effectiveness and efficiency of drug discovery, which continues along the one-
drug-one-gene paradigm. Consequently, the cost of bringing a drug to market is staggering, and the failure rate
is daunting. Our long-term goal is to revive the lagging pharmaceutical pipeline by identifying robust methods
for achieving precision medicine. We will achieve this goal by developing a novel structural systems
pharmacology approach to drug discovery, which integrates structure-based drug design with heterogeneous
omics data integration and analysis in the context of the whole human and pathogen genome and interactome.
An increasing body of evidence from both our group and others suggests that most drugs commonly interact
with multiple receptors (targets). Both strong and weak multiple drug-target interactions can collectively
mediate drug efficacy, toxicity, and resistance through the conformational dynamics of biomolecules. In order
to rationally design potent, safe, and precision medicine, we face one of the major unsolved challenges in
structure-based drug design: what are all the possible proteins and their conformational states interacting with
a drug in an organism? This proposal attempts to address this challenge by developing, disseminating, and
experimentally testing novel computational tools. Based on our successful preliminary results, we will develop
an integrated computational pipeline to identify three-dimensional (3D) protein-chemical interaction models in
the cellular context and on a structural proteome scale. Specifically, we will develop a quaternary structure-
centric multi-layered network model by integrating heterogeneous data from genomics, proteomics, and
phenomics. We will develop a novel collaborative one-class collaborative filtering algorithm to infer missing
relations in the multi-layered network. We will combine tools derived from structural bioinformatics, biophysics,
and machine learning to gain biological insights into the drug action. To facilitate the usability and
reproducibility of the proposed algorithms, we will develop community-based web resources established by our
previous experiences in developing the Protein Data Bank (PDB). More importantly, we will work closely with
experimental laboratories to test the proposed computational tools using targeted kinase polypharmacology as
a real-world example, and iteratively improve the performance and usability of algorithm, software, and web
services. The successful completion of this project will provide the scientific community with: (1) new methods
to enhance the scope and capability of high-throughput screening for structure-based multi-target drug design;
(2) a user-friendly web service to support community-based drug discovery; and (3) potential novel anti-cancer
targeted therapeutics. Together, these tools will advance drug discovery and precision medicine by providing a
structural systems pharmacology toolkit.