Seeing Change in a New Light: Application of Reversible Jump MCMC for Changepoint Detection in Single Molecule Imaging

Description

Lung cancer is one of the leading causes of cancer deaths worldwide with Non-Small Cell Lung Cancer (NSCLC) accounting for over 80% of all lung cancer diagnoses (Cancer.net, 2024). The development of NSCLC is in part due to mutations in the DNA of Epidermal Growth Factor Receptors (EGFRs); proteins involved in normal cell growth and development. Combinations of mutations can trigger difficult to predict changes in clustering of the EGFR on the cell membrane. Measuring separations between EGFR proteins in these clusters can provide a better understanding of the underlying molecular processes leading to cancer development. This understanding can enable more rapid identification of novel treatments and more precise targeting of existing treatments through the development of diagnostic tests to stratify cancer patients according to their EGFR clustering profile.

Fluorescence Localisation Imaging with Photobleaching (FLImP) (Iyer et. al., 2023) is a single molecule imaging technique capable of resolving fluorescently labelled EGFR molecules. Fluorophores, bound to the EGFRs on the cell surface, emit light which can be measured over time. However, due to the proximity of the fluorophores, the emitted intensities’ point spread functions (PSFs) overlap, creating a single spot of their combined intensities alongside background noise, complicating individual identification. To overcome this, FLImP relies on the sequential photobleaching of fluorophores in close proximity. This process creates a staircase-like pattern in the one-dimensional fluorophore intensity profile over time. Modelling the number and location of photobleaching events in these one-dimensional intensity profiles can be approached as a changepoint problem and can be used to estimate the number of fluorophores active in each observation frame. This information, combined with knowledge about the shape of the PSFs, can then be used to determine the separations between the EGFR molecules to which the fluorophores are bound.

The research presented outlines the use of reversible jump Markov chain Monte Carlo applied to changepoint problems such as these to effectively and efficiently model the number of active fluorophores present in each frame (Green, 1995). We apply this method to both simulated data and FLImP data for comprehensive validation and analysis, and provide a comparison to other methods within the field. Avenues for future research within this area will also be discussed.

“Lung Cancer – Non Small Cell: Statistics.” Cancer.Net, 29 Feb. 2024, https://www.cancer.net/cancer-types/lung-cancer-non-small-cell/statistics.

Iyer, R. Sumanth, et al. “The T766M-EGFR lung cancer mutation promotes tumor growth by exploiting newfound mechanisms assembling ligand-free EGFR oligomer structures.” bioRxiv (2023) 2023-06.

Green, Peter J. "Reversible jump Markov chain Monte Carlo computation and Bayesian model determination." Biometrika 82.4 (1995): 711-732.

Period	15 May 2024
Event title	44th Conference on Applied Statistics in Ireland
Event type	Conference
Conference number	44
Location	Athlone, IrelandShow on map
Degree of Recognition	National