University of British Columbia
Simulation-based inference of micelle geometry from 2D cryo-EM images of membrane proteins
Abstract
dc:descriptionMembrane proteins are important pharmaceutical targets, but the intrinsic hydrophobic nature of their transmembrane domains imposes a challenge for experimental structural studies. The inclusion of membrane-mimetic detergents in the extraction stabilizes the protein but adds a new factor to consider in structural experiments. Although cryo-EM is a revolutionizing tool, there is still a lack of methods to properly disentangle the contribution of the protein from the detergent micelle to the generated images. In this work, we review the properties and availabilities of micelles images from cryo-EM studies of membrane proteins and outline the development of an in-silico model of the protein-detergent system and a pipeline for inferring geometric parameters of micelles from cryo-EM simulated images. Since our inference method relies on the Approximation Bayesian Computation (ABC) algorithm, we tested different scoring metrics and ran the method on simulated images from existing structures. As our inferring algorithm yields a distribution of micelle parameters that approximates well the ground truth value in our experiments, we obtain a proof-of-concept that will need more refinements to get applied to experimental data but also hold some promising applications for automating and improving cryo-EM imaging of membrane proteins.
Degree
thesis:*- Name thesis:degree_name
- Master of Science - MSc
- Level thesis:degree_level
- master's
- Discipline thesis:degree_discipline
- Bioinformatics
- Grantor dc:publisher
- University of British Columbia
- Year dc:date
- 2023
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- de Castro Alvarenga, Helena
- Contributors dc:contributor
-
- Dao Duc, Khanh
Rights
dc:rights- Statement dc:rights
-
- Attribution-NonCommercial-NoDerivatives 4.0 International
- Language dc:language
- eng
Identifiers
dc:identifier.*- Handle dc:identifier
- http://hdl.handle.net/2429/87138