Applications of Boolean modelling to study and stratify dynamics of a complex disease

Interpretation of omics data is needed to form meaningful hypotheses about
disease mechanisms. Pathway databases give an overview of disease-related processes, while mathematical models give qualitative and quantitative insights into
their complexity. Similarly to pathway databases, mathematical models are stored
and shared on dedicated platforms. Moreover, community-driven initiatives such
as disease maps encode disease-specific mechanisms in both computable and
diagrammatic form using dedicated tools for diagram biocuration and visualisation. To investigate the dynamic properties of complex disease mechanisms,
computationally readable content can be used as a scaffold for building dynamic
models in an automated fashion. The dynamic properties of a disease are extremely complex. Therefore, more research is required to better understand the
complexity of molecular mechanisms, which may advance personalized medicine
in the future.
In this study, Parkinson’s disease (PD) is analyzed as an example of a complex
disorder. PD is associated with complex genetic, environmental causes and
comorbidities that need to be analysed in a systematic way to better understand
the progression of different disease subtypes. Studying PD as a multifactorial
disease requires deconvoluting the multiple and overlapping changes to identify
the driving neurodegenerative mechanisms. Integrated systems analysis and
modelling can enable us to study different aspects of a disease such as progression,
diagnosis, and response to therapeutics. Therefore, more research is required to
better understand the complexity of molecular mechanisms, which may advance
personalized medicine in the future. Modelling such complex processes depends
on the scope and it may vary depending on the nature of the process (e.g. signalling
vs metabolic). Experimental design and the resulting data also influence model
structure and analysis. Boolean modelling is proposed to analyse the complexity
of PD mechanisms. Boolean models (BMs) are qualitative rather than quantitative
and do not require detailed kinetic information such as Petri nets or Ordinary
Differential equations (ODEs). Boolean modelling represents a logical formalism
where available variables have binary values of one (ON) or zero (OFF), making it
a plausible approach in cases where quantitative details and kinetic parameters
9
are not available. Boolean modelling is well validated in clinical and translational
medicine research.
In this project, the PD map was translated into BMs in an automated fashion
using different methods. Therefore, the complexity of disease pathways can be
analysed by simulating the effect of genomic burden on omics data. In order to
make sure that BMs accurately represent the biological system, validation was
performed by simulating models at different scales of complexity. The behaviour
of the models was compared with expected behavior based on validated biological
knowledge. The TCA cycle was used as an example of a well-studied simple
network. Different scales of complex signalling networks were used including the
Wnt-PI3k/AKT pathway, and T-cell differentiation models. As a result, matched
and mismatched behaviours were identified, allowing the models to be modified
to better represent disease mechanisms. The BMs were stratified by integrating
omics data from multiple disease cohorts. The miRNA datasets from the Parkinson’s Progression Markers Initiative study (PPMI) were analysed. PPMI provides
an important resource for the investigation of potential biomarkers and therapeutic targets for PD. Such stratification allowed studying disease heterogeneity and
specific responses to molecular perturbations. The results can support research
hypotheses, diagnose a condition, and maximize the benefit of a treatment. Furthermore, the challenges and limitations associated with Boolean modelling in
general were discussed, as well as those specific to the current study.
Based on the results, there are different ways to improve Boolean modelling
applications. Modellers can perform exploratory investigations, gathering the
associated information about the model from literature and data resources. The
missing details can be inferred by integrating omics data, which identifies missing
components and optimises model accuracy. Accurate and computable models
improve the efficiency of simulations and the resulting analysis of their controllability. In parallel, the maintenance of model repositories and the sharing of
models in easily interoperable formats are also important.