Project Overview
This MPhys project studies how chromatin organisation can influence transcriptional dynamics and cellular identity through the language of Waddington’s landscape. Using a coarse-grained polymer model of bridging-induced phase separation (BIPS), chromatin is represented as a bead–spring fibre and transcription factors as diffusing multivalent binders that can transiently bridge distant regions of the genome. This provides a physical framework for understanding how three-dimensional chromatin architecture may help stabilise developmental states.
From the simulated chromatin–transcription factor dynamics, I constructed binary transcription-unit activity traces and analysed them using a telegraph model. This allowed me to extract effective switching rates, burst statistics, and correlation times, and to compare stem-like, differentiated, and senescent-like chromatin states. The main result was that differentiation redistributes activity away from flexible intermediate switching regimes towards more strongly locked ON and OFF states, suppresses burst initiation, and concentrates transcription into a smaller subset of dominant units — a process I describe as fossilisation.
I then followed differentiation dynamically using both chromatin-remodelling and transcription-factor-collapse protocols, showing that commitment emerges primarily through suppression of OFF→ON reactivation rather than a simple global shutdown of activity. Finally, reversal simulations revealed that transcription-factor abundance strengthens memory and canalisation: higher TF number makes differentiated transcriptional programmes more persistent and more attractor-like during reversal.
What I Learned
This project gave me extensive experience in computational biophysics, polymer modelling, and transcriptional dynamics. I worked with LAMMPS and VMD to simulate chromatin folding, translated and extended baseline simulation code, and developed new dynamical protocols to model differentiation and reversal. I also built a Python analysis pipeline to extract transcriptional observables from simulation output, including telegraph-model inference, burst statistics, heterogeneity, plasticity, programme drift, and correlation times.
More broadly, the project taught me how to connect abstract biological ideas — such as differentiation, commitment, memory, and Waddington’s landscape — to explicit physical observables in a controlled model. It deepened my understanding of chromatin physics, stochastic processes, and soft matter approaches to gene regulation, while also strengthening my scientific writing, figure design, and the ability to turn a large simulation project into a coherent research story.