{"id":{"repo_id":"cambridge","oai_identifier":"oai:www.repository.cam.ac.uk:1810/390412"},"canonical_url":"https://search.dev.ndltd.org/etd/cambridge/oai:www.repository.cam.ac.uk:1810/390412","repository":{"repo_id":"cambridge","name":"Cambridge University","base_url":"https://api.repository.cam.ac.uk/server/oai/request"},"display":{"title":"Machine learning for cryo-EM structure determination","abstract":"Electron cryo-microscopy (cryo-EM) has revolutionised structural biology thanks to its relaxed sample requirements. Recent advances in cryo-EM image processing have enabled the structural determination of many macromolecules that were previously intractable by other methods. At the same time, breakthroughs in machine learning—most notably AlphaFold2—have driven further progress across structural biology. During my PhD, I developed novel machine-learning methods to enhance cryo-EM data processing. In single-particle cryo-EM, thousands of 2D projections of a macromolecule are collected in unknown orientations. Software such as RELION reconstructs these projections into a 3D electron density map, which must then be interpreted as an atomic model—a manual process that can take months for large complexes. My doctoral work focused on ModelAngelo, an automated model-building and protein-identification program for cryo-EM. I designed specialised graph-neural-network architectures that allow ModelAngelo to build atomic models in high-resolution maps with accuracy matching that of human experts. I also developed a sequence-free version of ModelAngelo to enable protein identification when some complex components are unknown; this tool has already seen widespread adoption, with several applications described in this thesis. Cryo-EM datasets often include macromolecules in heterogeneous conformational states. Variational auto-encoders (VAEs) have recently been applied to disentangle these conformations by encoding maps into a low-dimensional latent space. As an additional part of my doctoral work, in a collaboration with Dari Kimanius, I developed a VAE with a novel decoder architecture that efficiently resolves cryo-EM heterogeneity via linear combinations of learned volumes.","abstract_html":"Electron cryo-microscopy (cryo-EM) has revolutionised structural biology thanks to its relaxed sample requirements. Recent advances in cryo-EM image processing have enabled the structural determination of many macromolecules that were previously intractable by other methods. At the same time, breakthroughs in machine learning—most notably AlphaFold2—have driven further progress across structural biology. During my PhD, I developed novel machine-learning methods to enhance cryo-EM data processing. In single-particle cryo-EM, thousands of 2D projections of a macromolecule are collected in unknown orientations. Software such as RELION reconstructs these projections into a 3D electron density map, which must then be interpreted as an atomic model—a manual process that can take months for large complexes. My doctoral work focused on ModelAngelo, an automated model-building and protein-identification program for cryo-EM. I designed specialised graph-neural-network architectures that allow ModelAngelo to build atomic models in high-resolution maps with accuracy matching that of human experts. I also developed a sequence-free version of ModelAngelo to enable protein identification when some complex components are unknown; this tool has already seen widespread adoption, with several applications described in this thesis. Cryo-EM datasets often include macromolecules in heterogeneous conformational states. Variational auto-encoders (VAEs) have recently been applied to disentangle these conformations by encoding maps into a low-dimensional latent space. As an additional part of my doctoral work, in a collaboration with Dari Kimanius, I developed a VAE with a novel decoder architecture that efficiently resolves cryo-EM heterogeneity via linear combinations of learned volumes.","abstract_has_math":false,"creators":["Jamali, Kiarash"],"institution":"University of Cambridge","degree_name":"Doctor of Philosophy (PhD)","degree_level":"Doctoral","degree_discipline":null,"degree_department":null,"school":null,"contributors":[],"advisors":["Scheres, Sjors HW"],"committee_chairs":[],"committee_members":[],"year":2025,"date_issued":"2025-05-16","date_published":"2025-05-16","updated_at":"2026-07-22T22:24:13Z","subjects":["cryo-EM","Machine learning","protein structure"],"languages":["eng"],"rights":[],"rights_urls":["https://www.repository.cam.ac.uk/bitstreams/8bb552d8-1ba2-4c8b-b939-b58bd8f3bf7b/download","http://purl.org/NET/rdflicense/allrightsreserved"],"identifier_entries":[]},"links":{"outbound_url":"https://doi.org/10.17863/CAM.121976","outbound_label":"DOI","outbound_source":"dc:identifier.doi"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor.advisor","label":"Advisor","values":["Scheres, Sjors HW"]},{"key":"dc:contributor.sponsor","label":"Sponsor","values":["Cambridge Trust"]},{"key":"dc:creator","label":"Author","values":["Jamali, Kiarash"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.issued","label":"Date","values":["2025-05-16"]},{"key":"dc:publisher.institution","label":"Dc Publisher Institution","values":["University of Cambridge"]},{"key":"dc:relation.isreferencedby.uri","label":"Dc Relation Isreferencedby URI","values":["https://www.repository.cam.ac.uk/handle/1810/390412"]},{"key":"dc:type","label":"Dc Type","values":["Thesis"]},{"key":"dc:type.qualificationlevel","label":"Dc Type Qualificationlevel","values":["Doctoral"]},{"key":"dc:type.qualificationname","label":"Dc Type Qualificationname","values":["Doctor of Philosophy (PhD)"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["cryo-EM","Machine learning","protein structure"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language","label":"Dc Language","values":["eng"]},{"key":"dc:rights","label":"Dc Rights","values":["https://www.repository.cam.ac.uk/bitstreams/8bb552d8-1ba2-4c8b-b939-b58bd8f3bf7b/download","http://purl.org/NET/rdflicense/allrightsreserved"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier.doi","label":"DOI","values":["https://doi.org/10.17863/CAM.121976"]},{"key":"dc:identifier.uri","label":"Identifier URI","values":["https://www.repository.cam.ac.uk/bitstreams/2798944e-9671-4865-abe5-24c39ed9c94d/download"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["Electron cryo-microscopy (cryo-EM) has revolutionised structural biology thanks to its relaxed sample requirements. Recent advances in cryo-EM image processing have enabled the structural determination of many macromolecules that were previously intractable by other methods. At the same time, breakthroughs in machine learning—most notably AlphaFold2—have driven further progress across structural biology. During my PhD, I developed novel machine-learning methods to enhance cryo-EM data processing. In single-particle cryo-EM, thousands of 2D projections of a macromolecule are collected in unknown orientations. Software such as RELION reconstructs these projections into a 3D electron density map, which must then be interpreted as an atomic model—a manual process that can take months for large complexes. My doctoral work focused on ModelAngelo, an automated model-building and protein-identification program for cryo-EM. I designed specialised graph-neural-network architectures that allow ModelAngelo to build atomic models in high-resolution maps with accuracy matching that of human experts. I also developed a sequence-free version of ModelAngelo to enable protein identification when some complex components are unknown; this tool has already seen widespread adoption, with several applications described in this thesis. Cryo-EM datasets often include macromolecules in heterogeneous conformational states. Variational auto-encoders (VAEs) have recently been applied to disentangle these conformations by encoding maps into a low-dimensional latent space. As an additional part of my doctoral work, in a collaboration with Dari Kimanius, I developed a VAE with a novel decoder architecture that efficiently resolves cryo-EM heterogeneity via linear combinations of learned volumes."]},{"key":"dc:format.checksum.md5","label":"Dc Format Checksum Md5","values":["bc509eeb0f8b0094392cc3b12b04ef72","87eda9de84448d1f82354d60eee3eb5f"]},{"key":"dc:title","label":"Title","values":["Machine learning for cryo-EM structure determination"]}]}],"canonical_facts":{"dc:contributor.advisor":["Scheres, Sjors HW"],"dc:contributor.sponsor":["Cambridge Trust"],"dc:creator":["Jamali, Kiarash"],"dc:date.issued":["2025-05-16"],"dc:description.abstract":["Electron cryo-microscopy (cryo-EM) has revolutionised structural biology thanks to its relaxed sample requirements. Recent advances in cryo-EM image processing have enabled the structural determination of many macromolecules that were previously intractable by other methods. At the same time, breakthroughs in machine learning—most notably AlphaFold2—have driven further progress across structural biology. During my PhD, I developed novel machine-learning methods to enhance cryo-EM data processing. In single-particle cryo-EM, thousands of 2D projections of a macromolecule are collected in unknown orientations. Software such as RELION reconstructs these projections into a 3D electron density map, which must then be interpreted as an atomic model—a manual process that can take months for large complexes. My doctoral work focused on ModelAngelo, an automated model-building and protein-identification program for cryo-EM. I designed specialised graph-neural-network architectures that allow ModelAngelo to build atomic models in high-resolution maps with accuracy matching that of human experts. I also developed a sequence-free version of ModelAngelo to enable protein identification when some complex components are unknown; this tool has already seen widespread adoption, with several applications described in this thesis. Cryo-EM datasets often include macromolecules in heterogeneous conformational states. Variational auto-encoders (VAEs) have recently been applied to disentangle these conformations by encoding maps into a low-dimensional latent space. As an additional part of my doctoral work, in a collaboration with Dari Kimanius, I developed a VAE with a novel decoder architecture that efficiently resolves cryo-EM heterogeneity via linear combinations of learned volumes."],"dc:format.checksum.md5":["bc509eeb0f8b0094392cc3b12b04ef72","87eda9de84448d1f82354d60eee3eb5f"],"dc:identifier.doi":["https://doi.org/10.17863/CAM.121976"],"dc:identifier.uri":["https://www.repository.cam.ac.uk/bitstreams/2798944e-9671-4865-abe5-24c39ed9c94d/download"],"dc:language":["eng"],"dc:publisher.institution":["University of Cambridge"],"dc:relation.isreferencedby.uri":["https://www.repository.cam.ac.uk/handle/1810/390412"],"dc:rights":["https://www.repository.cam.ac.uk/bitstreams/8bb552d8-1ba2-4c8b-b939-b58bd8f3bf7b/download","http://purl.org/NET/rdflicense/allrightsreserved"],"dc:subject":["cryo-EM","Machine learning","protein structure"],"dc:title":["Machine learning for cryo-EM structure determination"],"dc:type":["Thesis"],"dc:type.qualificationlevel":["Doctoral"],"dc:type.qualificationname":["Doctor of Philosophy (PhD)"]},"updated_at":"2026-07-22T22:24:13Z"}