2D matching with 3D template

In cryo-EM, a single particle is captured using an electron microscope. Therefore, we obtain a large number of 2D projections of a 3D molecule. Typically, the number of projections can be up to 100,000. The automatic detection of these molecules from the acquired micrograph is a primary challenge that still leads to recent development. The task can be framed as the detection of 2D pattern corresponding to a same 3D structure.

A similar and probably most interesting setting arises also in cryo-ET. In cryo-ET, a 3D object (e.g. an entire cell) is imaged by a electron microscope while tilting the 3D object. This results in a set of 2D projections also called a tilt series. Among what constitutes the cell, some small structures such as ribosomes are present a very large number of times. Standard post-processing techniques in cryo-ET as [1] use this repetition to combine all the information of these smaller structures into a better resolved image (or at least a better resolved ribosome). This problem can be again cast as the detection of 2D patern corresponding to a same 3D structure.

In this project, we will investigate original techniques combining standard signal processing and new machine learning methods to improve existing methods in 2D matching with 3D templates.

This project requires a intermediate coding skill and a strong interest for solving practical problems. Basic knowledge of signal processing and or signal processing is required.

This project can be adapted into either a Bachelor or a Master thesis.

You will be supervised by at least two different person. Please reach out with one of the following for a first inquiry:

• Valentin Debarnot, valentin.debarnot[@]unibas.ch
• Ivan Dokmanić, ivan.dokmanic[@]unibas.ch

[1] Bronwyn A Lucas, Benjamin A Himes, Liang Xue, Timothy Grant, Julia Mahamid, and Nikolaus Grigorieff. Locating macromolecular assemblies in cells by 2d template matching with cistem. Elife, 10, 2021