Machine learning for finding macromolecules in cells
Medical Research Council (Cambridge)
Not stated
- Funding
- Competition Funded PhD Project (Students Worldwide)
- Application deadline
- 8 December 2026
About the project
About the Project The aim of this PhD project is to develop computer software for in situ structural biology projects. Specifically, software tools for the identification of macromolecular complexes in cryo-EM/cryo-ET data sets collected on thinned human and budding yeast cells. The main target is the kinetochore , but the methods developed will have applicability to other rare complexes present in cells. The research in my group is focused on understanding the mechanisms and regulation of chromosome segregation in mitosis. During the cell cycle, accurate chromosome segregation ensures that both daughter cells inherit the correct complement of chromosomes. Errors in this process cause aneuploidy leading to cancer and developmental defects. Duplicated chromosome are segregated in mitosis by the mitotic spindle . Each chromosome is attached to microtubules by kinetochores, large protein complexes that specifically assemble onto centromeric chromatin. Kinetochores that mediate and regulate this process consist of over 100 proteins, that also function to detect and signal appropriate microtubule attachment and tension. Previously we have used single particle cryo-EM to determine structures of kinetochores. Our aim now is to visualise kinetochores in situ. For this we are applying cryo-electron tomography and zero-tilt cryo-EM to thinned human and yeast cells. We have established work-flows for targeting kinetochores for data collection and generation of cryo-electron tomograms. Identifying kinetochores in the dense chromatin structure of the nucleus is difficult, and current 3-dimensional template matching programs have failed. In this PhD project we aim to apply machine learning and AI tools to identify kinetochores in our cryo-EM/ET data sets. We will use cryo-EM/ET data with and without kinetochores for training. The PhD applicants should have a good understanding of AI and machine learning tools, be proficient in computer programing (eg Python), and be interested in structural biological questions.