CellFrame: A Data Structure for Abstraction of Cell Biology Experiments and Construction of Perturbation Networks

Abstract

Different cell types could respond to the same set of stimuli in very different ways, so it is important to collect and integrate these experimental data in a cell-type specific manner in order to properly model these processes. In practice, however, cellular or biochemical models were usually constructed separately, with data from multiple cell sources, making it difficult to compare or combine these models. To circumvent this problem, we propose to conduct the model integration at the data level. To facilitate this purpose we have designed a data structure (CellFrame) to store cell-specific data, such as cellular component measurement and stimulus-response, and to infer cell perturbation network (a cellular network representing the stimulus-response relationships between the cellular and/or environmental components). The CellFrame database consists of two data types: data classes and supporting classes. Data classes constitute three modules: cell component measurement, qualitative cell perturbation, and quantitative cell perturbation. Supporting classes serve as adaptors, linking to external molecular or literature databases. We have implemented an initial version of CellFrame, which contains data collected from reported experiments on human brain astrocytoma and colorectal cancer cell lines (http://cellframe.bioknowledge.org). Perturbation networks are inferred following Boolean differential calculus. CellFrame will provide an opportunity to integrate cell biological data from a wide variety of experimental groups to help build cell models and design further experiments.