Boolean Networks for Genetic Pathway Modelling

This project aims to simulate protein-protein interactions within single cells of lung adenocarcinoma, in order to predict morphological phenotypes from mRNA and mutation data. Whilst many studies use ODE modelling, we opt for describing the transcription network as Boolean, allowing for more scalability and generality. This technique binarizes each species in the model by defining a threshold level it can either be above (ON) or below (OFF). We aim to construct a protein interaction network from both qualitative and quantitative sources, and then use a reverse-engineering approach to train the network to time-series perturbation data.

Once constructed, we aim to run stochastic simulations under different environmental conditions and mutation perturbations, to get probability estimates as to the phenotype of the cell (i.e. the stable state of the simulated model). Adding such random elements into the simulations allows multiple runs of a single cell model to represent a heterogeneous cluster of similar cells.

Finally patient sample data will be inputted into the model, including mRNA expression levels, CNAs and specific mutations unique to each patient. These will define the parameters and initial values of our model simulations, resulting in the simulated cell phenotypes being our prediction for the physiological progression of the tumour. Also, we aim to incorporate multiplex, spatial, single-cell protein data from Cancer Research UK to more accurately reflect the heterogeneity of the model