Mart So-Last
Postdoc
I’m Mart So-Last, postdoc in Arjen Jakobi’s group at TU Delft. My work revolves around cellular cryo-ET and combines method development with biological applications. During my PhD in Leiden, we focused on the development of correlative single-molecule localization microscopy and cryo-ET. The idea was to localize particles of interest with high precision via light microscopy, to then ensure we could image these with cryo-ET. This involved labelling: we would label a structure in the cell, in order to find it in our sample. We also developed a software package, scNodes, to help with the fairly laborious data processing involved in this correlative approach.
Along the way I became interested in a different approach to labelling. Rather than relying on correlative microscopy, I became more interested in workflows that combined large-scale cryo-ET data acquisition with machine learning for data processing, using software such as Ais and Pom. Instead of labelling a structure in our sample, we could try to label or detect features of interest directly in the data.
I obtained an EMBO Postdoctoral Fellowship to apply this approach to the study of an organelle that has been hypothesized to exist in human cells, called the purinosome. During my postdoc at the MRC LMB in Cambridge, UK, with Matteo Allegretti, I developed easymode, a collection of pretrained general networks for feature detection in cellular cryo-ET datasets, and used it to determine the structure of IMPDH, the rate-limiting enzyme in purine synthesis that I believe to be one of the main components of purinosomes, within intact cells. In a separate project we also used this approach to identify and build an atomic model for the main microtubule inner protein in human neurons.
My current goal in Delft is to set up this same large-scale and mostly automated cellular cryo-ET workflow and to apply it to various problems - from gas vesicles to the molecular architecture of purinosomes.
