Watch our video presentation explaining our computational approach for using molecular data to determine the comparative biological impact of cigarette smoke and RRP aerosol.
Computational Biology
Computational Biology: Assessing biological impact qualitatively and quantitatively
There are four steps to the computational component of our systems toxicology approach to smoke-free product assessment.
Firstly, we generate Systems Response Profiles, which are a measure of the degree to which individual molecular entities are changed as a consequence of exposure to specific compounds.
Secondly, we build models of the biological processes relevant to smoking-related diseases, providing a layer of mechanistic understanding that allows us to go beyond the examination of single genes.
Thirdly, we calculate Network Perturbation Amplitudes (NPAs), which describe the magnitude of biological network alterations brought about by exposing tissue to a given stimulus.
Finally, NPAs are aggregated into the Biological Impact Factor representing the overall alteration of the biological system being studied as a result of exposure to smoke-free product aerosols, exposure to cigarette smoke or smoking cessation.
Learn more about:
Network Construction
Biological Network Construction
The biological network models that we build relate to the cardiovascular and pulmonary contexts, the two tissues that are highly impacted by exposure to cigarette smoke. Built using the Biological Expression Language (BEL), that allows the presentation of biological processes in a computable and human-readable format, they combine a variety of biological pathways to represent important processes such as cell proliferation, cellular stress, cell fate, inflammation and tissue repair.
Constructing biological network models
Biological network models are built respecting, as much as possible, the biological boundaries of both the biological process to be modelled and the tissue / cell type of interest. There are three main steps to the process.
Literature backbone
Scientific literature within the boundaries is reviewed and cause and effect relationships are encoded in BEL. Put together, these relationships form a literature backbone consisting of nodes (the entity of interest, eg, a protein or gene) and edges (the causal relationships between the nodes).
Model enhancement
Using Reverse Causal Reasoning (RCR), gene expression data from biologically relevant experiments are scored against a BEL statement knowledge base to identify biological entities that are predicted to be impacted. Through this process, nodes in the literature backbone are verified and the network models enhanced with additional biology.
Validation
The integrated network models are then validated using RCR and molecular data from experiments that demonstrate the impact on the specific biological processes that the models represent.
Our biological network models
Using this approach, we have to date built the following biological networks for Reduced-Risk Product assessment:
A set of biological networks also exists for essential pathobiological events in vascular disease[6] and we have further augmented the models initially built in the non-diseased pulmonary context with biology relevant to chronic obstructive pulmonary disease[7].
Despite the systematic internal verification of biological network models described above, our models are further subjected to crowd-verification as part of the sbv IMPROVER project.
References:
NPAs and the BIF
Network Perturbation Amplitudes (NPAs) describe the magnitude of biological network alterations brought about by exposing tissue to a given stimulus (for example, cigarette smoke or smoke-free product aerosols). They are achieved by applying algorithms to high-throughput transcriptomic data and relevant biological network models[1].
In a set of networks, NPA scores can be calculated for each and then aggregated into one holistic score, the Biological Impact Factor (BIF), which represents the measure of the overall perturbation of the biological system being studied[2][3].
We have shown the applicability of the BIF score as a predictor for medium- and long-term disease outcome by using the transcriptomic data from nasal epithelium of rats following exposure to multiple doses of formaldehyde for 13 weeks[4]. In the original study, the formaldehyde dose correlated with the tumorigenesis rate after two years[5] and we were able to demonstrate an excellent correlation between the BIF scores computed from the transcriptomic data and tumorigenesis rates.
Graph showing tumorigenesis rates and BIF scores.
Note: These data alone do not imply or represent a claim of reduced exposure or reduced risk.
Not only a predictive tool for toxic outcome, our BIF approach also provides insight into the biological mechanisms that potentially link exposure to disease risk. We are able to apply BIF scoring in the comparison of harm associated with cigarettes and smoke-free products, as well as in the evaluation of any unwanted biological impact resulting from smoke-free products.
