SMBp Seminar: Kexin Zhang (UMass)

Monday Oct 2, 2023, 2:00 PM → 3:00 PM America/New_York

3rd Floor Classroom/3-Flatiron Institute (162 5th Avenue)

Speaker: Kexin Zhang (UMass)

 

Topic: Exploring the limits of 2D template matching for detecting targets in cellular cryoEM images

 

Abstract: Accurately placing macromolecular assemblies in the cellular context is an important step in
understanding their mechanistic role inside the cell. Previously, our lab developed a 2D
template matching (2DTM) approach to detect targets in cellular cryoEM images with high
positional and orientational accuracy. In 2DTM, we calculate a signal-to-noise ratio (SNR) for
every location in the cryoEM image that depends on the agreement between the template and
the image. A target is detected when the SNR value exceeds a statistically defined threshold.
2DTM offers a way to study macromolecular assemblies in a broader context of a cell, taking
advantage of the increasing number of available high-resolution structures as templates.
Building on the success in locating and discriminating larger molecular species in cells, we seek
to improve 2DTM to detect more challenging targets, such as rare and relatively small
complexes. In our current work, we show that the autocorrelation of the noise-free templates
provides complementary information to the cross-correlation-based 2DTM SNR. By comparing
the cross-correlation and autocorrelation scores, we can better distinguish false detection from
true detection. We quantify this comparison as another correlation evaluation and develop a
probabilistic model of target detection via 2DTM in cellular cryoEM images.