In this new era of Big Data, we can leverage the abundance of large Omics datasets: high-throughput biochemical assays that measure concentration and availability in cells such as genomics profile DNA, transcriptomics measure transcripts; proteomics and metabolomics quantify proteins and metabolites respectively.
We can leverage this availability to find new drugs, targets, and populations to help optimize the treatment of disease. Cataloging the genomic, transcriptomic, proteomic, and metabolomic changes in a system using new high-throughput screening is essential in understanding multifactorial influences that lead to disease.
However, we are reaching the end of an era where we can find a single magic bullet for a single target to treat a disease. Now, we need to understand how genes, proteins, small molecules, and metabolites interact to create an integrated model for disease progression. In addition to finding individual biomarkers for drug therapy, we can also discover the systems-level rules that connect biological networks to their resulting phenotypes. Understanding the rules that govern biological networks can open the door to drug repurposing, combination therapies, population optimization, and improved in-silico models of drug efficacy and safety, potentially saving millions of dollars on failed clinical trials. This also speeds up the overall drug discovery process as well. We propose that by taking this new integrated approach, we can revolutionize pharmaceutical discovery.
To know the true causes of disease we need to couple systems biology and network analysis in a framework that we call "Network Medicine". Network Medicine integrates information from complicated Omics platforms with applied mathematics to analyze the rules that combine the structure of biological networks and their resultant effect on health. The generation and analysis of chemical-biological networks can form integrative models to infer and assess how drugs affect biological function. This can greatly enhance the prioritization of drug candidates for further investigation.
The field of Network Medicine integrates research from many disciplines including chemistry, genomics, molecular biology, statistical physics, biostatistics, and bioinformatics. In addition, Network Medicine requires different multiscale and time-dependent datasets to be integrated and analyzed using advanced knowledge of data storage and analysis including mastery of natural language processing (NLP), machine learning, and dynamic simulations. As the field is inherently multidisciplinary, no single investigator can completely master all the research areas. Therefore, interdisciplinary groups in pharmaceutical industries with individuals from various functions and backgrounds will collaborate to use an integrative approach to take the knowledge gained from biological screening into a holistic understanding of the biology underlying health and disease.
A network is defined as a series of entities, known as nodes, connected to one another on the basis of a defined criterion, known as edges. Nodes represent variables such as the abundance and state of genes, proteins, drugs, metabolites, and diseases. Nodes in a network can also be used to specify the state of a system. The connections between the nodes are termed edges and can be specified using criteria of interest. In the context of studying drug action, edges may represent protein-protein interactions or drug-target interactions. Edges can also be defined on the basis of overlapping properties between two nodes such as the chemical-binding, disease-association, pathway-membership, genetic, or structural similarity. These types of complex definitions for edges allow the networks to incorporate multiscale interactions, from microscopic molecular-level drug-target interactions to the coordinated functional outputs of multiple organs, i.e., phenotype.
Simulations of how phenotypes are formed by the structure of a biological network over time can be computed using Boolean dynamics. In Boolean dynamics, each node has a chance to exist in two states (inactive or active) and each state has defined rules on how it influences the state of other nodes. Useful dynamic models have also been created using ordinary differential equations to simulate how quantitative variables such as gene expression and drug efficacy change over time. We propose to incorporate bleeding-edge advances in network science, such as agent-based modeling and fuzzy logic, to improve how Network Medicine can be used to facilitate pharmaceutical development.
The Human Interactome
The human interactome currently encompasses ≈25,000 protein-coding genes, ≈20,000 metabolites, and an ever-increasing number of post-translationally modified proteins and functional-RNA molecules. Together, this exceeds 100K participants and a massive amount of potential functional interactions. To bring meaning to this large search space, we can quantify properties of the whole network. For example, the human interactome displays significant clustering, with pockets of especially dense interconnectedness. The human interactome has a “scale-free” behavior, defined as a network where interactions do not occur randomly, but rather follow a degree distribution with a power law. These networks have a few highly connected hubs and handle maintenance of the robustness of the network. Hub proteins are often evolutionarily conserved proteins which control many phenotypic outcomes.
The following has been hypothesized concerning biological networks:
- Non-essential genes move to the periphery of the interactome
- Proteins involved in the same disease tend to interact with each other
- Mutations in interacting proteins lead to similar phenotypes
- Diseases that share disease-associated cellular components show phenotypic similarity
By applying these hypotheses to the human interactome, we can find new purposes for FDA-approved drugs, find effective drug combinations, and improve predictions on the efficacy and safety of drug candidates.
Challenges of Network Medicine
Building a Network Medicine platform is not easy. There are several challenges that need to be resolved :
- The data lacks uniform ontologies. In order to integrate proprietary data with public datasets, we must connect diverse naming conventions such as disease, gene, protein, and compound IDs.
- The networks involved have heterogeneous data types. Many network algorithms were developed for homogenous data and must be adapted for multipartite networks.
- Biological networks are context- and time-dependent. For example, transcriptional responses of a drug in two cell lines can be completely disparate and integration of multiple models remains a complex challenge.
Because of the interconnected nature of its data, the biomedical domain has been one of the early adopters of graph databases, enabling more natural representation models and better data integration workflows, exploration, and analysis facilities. However, there are certain challenges such as the integration and searching of structural chemistry information.
How Katana Graph™ Can Help
The majority of platforms in the market today are struggling to keep up with the unprecedented scale and computing power required for predictive and AI analytics of vast volumes of complex biomedical data, synthesized from disparate sources.
At Katana Graph we understand these barriers and bottlenecks of Network Medicine and are committed to empowering users with the right platform to mine and analyze deep actionable insights from large complex datasets. Knowledge workers can now reliably store, query, mine, analyze, and develop AI models using heterogeneous data sources, to reveal breakthrough insights at scale and performance like never before. For instance, Katana Graph developed an integrated graph-RDKit cheminformatics platform that allows complex cheminformatics workflows to be streamlined across large, heterogeneous, biomedical knowledge graphs. This allows researchers to run a drug-disease association query to extract a subset of simplified molecular-input line-entry system (SMILES) representation of compounds. These researchers can then conduct specialized searches to identify similar chemical compounds that are in current clinical trials.
The Katana Graph Intelligence Platform is a distributed end-to-end platform designed to transform drug discovery processes. Regardless of the analyst's skill, Katana Graph’s AI-powered augmented graph analytics platform will facilitate decision intelligence and reduce the time to breakthrough innovations in healthcare and life sciences.