Humans are repeatedly exposed to many man-made chemicals, and the long-term effects of most of these chemical on human health are unknown. A starting point to improve our understanding is to determine how the addition of a chemical affects the processes that occur within cells. These processes are composed of tens of thousands of proteins and different types of interactions among them that form complex networks. Computational methods can analyze these networks in combination with experimental data on a chemical to determine the sub-network that is perturbed within the cell by that chemical. Such chemical response networks can be visualized in displays where each protein receives a location (x and y coordinates) such that proteins connected by an interaction are closely placed. There are many different ways to automatically assign these locations. Several software packages and web-based systems implement these graph layout algorithms. However, it is often very dif?cult to interpret these networks since the algorithms that determine these locations have very little knowledge of the underlying biology. Biologists often manually manipulate initial, automatically determined positions until the network con?guration matches their knowledge and intuition, but this process is time-consuming and requires signi?cant expertise. In this application, we seek to develop an online system that enables large, geographically distributed groups of non-experts (crowd users) to manipulate chemical response networks in order to make their visualized layouts meaningful to expert scientists. Our system will use automatically generated biological cues to guide the users on where to place nodes and edges, how to group them, and which nodes/edges to ignore. It will also enable users to highlight sequences of interactions that suggest how a chemical may control speci?c proteins and processes that occur in the cell. Our system will develop layouts through an iterative design process where users serve as barnstormers (generate several layout ideas), critics (offer principled feedback on layouts), and synthesizers (combine the best ideas from multiple layouts). We will employ three different types of crowd users: paid workers on Amazon Mechanical Turk, volunteer citizen scientists from Zooniverse, and college biology majors. We will perform extensive studies to evaluate the effectiveness of each group in creating graph layouts. We will also conduct a review by an expert panel of biomedical scientists to evaluate the effectiveness of our system in creating visualizations of chemical response networks that are valuable to experts and can assist them in formulating biological hypotheses.

Public Health Relevance

We will develop a web-based system called GraphSpace to enable citizen scientists and non-experts to improve visual representations of processes within the cell that are affected by environmental chemicals. GraphSpace will assist molecular biologists and toxicologists in discovering hypotheses about how these chemicals may harm human health.

Agency
National Institute of Health (NIH)
Institute
National Cancer Institute (NCI)
Type
Exploratory/Developmental Cooperative Agreement Phase I (UH2)
Project #
5UH2CA203768-02
Application #
9330822
Study Section
Special Emphasis Panel (ZRG1)
Program Officer
Miller, David J
Project Start
2016-08-11
Project End
2019-07-31
Budget Start
2017-08-01
Budget End
2019-07-31
Support Year
2
Fiscal Year
2017
Total Cost
Indirect Cost
Name
Virginia Polytechnic Institute and State University
Department
Biostatistics & Other Math Sci
Type
Biomed Engr/Col Engr/Engr Sta
DUNS #
003137015
City
Blacksburg
State
VA
Country
United States
Zip Code
24061
Bharadwaj, Aditya; Singh, Divit P; Ritz, Anna et al. (2017) GraphSpace: stimulating interdisciplinary collaborations in network biology. Bioinformatics 33:3134-3136