{"id":{"repo_id":"msu","oai_identifier":"oai:d.lib.msu.edu:etd_52671"},"canonical_url":"https://search.ndltd.org/etd/msu/oai:d.lib.msu.edu:etd_52671","repository":{"repo_id":"msu","name":"Michigan State University","base_url":"https://d.lib.msu.edu/oai"},"display":{"title":"From Task-Specific Tools to Foundation Models in Single-Cell Genomics","abstract":"Thesis (Ph.D.)--Michigan State University. Computer Science - Doctor of Philosophy, 2025","abstract_html":"Thesis (Ph.D.)--Michigan State University. Computer Science - Doctor of Philosophy, 2025","abstract_has_math":false,"creators":["Ding, Jiayuan"],"institution":null,"degree_name":null,"degree_level":null,"degree_discipline":null,"degree_department":null,"school":null,"contributors":["Tang, Jiliang","Qiu, Xiaojie","Xie, Yuying","Liu, Hui"],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2025,"date_issued":"2025","date_published":"2025","updated_at":"2026-11-18T18:59:26Z","subjects":["Artificial intelligence","Bioinformatics"],"languages":["eng"],"rights":["Attribution 4.0 International"],"rights_urls":[],"identifier_entries":[{"key":"dc:identifier","label":"Identifier","values":["etd:52671","local:Ding_grad.msu_0128D_21640"],"render_values":[{"text":"etd:52671","href":null,"code":true},{"text":"local:Ding_grad.msu_0128D_21640","href":null,"code":true}]}],"person_identifier_entries":[]},"links":{"outbound_url":"https://doi.org/doi:10.25335/3r45-1546","outbound_label":"DOI","outbound_source":"dc:identifier","record_url":"https://doi.org/doi:10.25335/3r45-1546","record_label":"Repository record","record_kind":"doi","record_source":"dc:identifier","download_url":null,"download_label":null,"download_source":null,"access_status":null,"embargo_until":null,"access_note":null},"source_record":{"url":"https://d.lib.msu.edu/oai?verb=GetRecord&metadataPrefix=oai_dc&identifier=oai%3Ad.lib.msu.edu%3Aetd_52671","prefix":"oai_dc"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor","label":"Contributor","values":["Tang, Jiliang","Qiu, Xiaojie","Xie, Yuying","Liu, Hui"]},{"key":"dc:creator","label":"Author","values":["Ding, Jiayuan"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date","label":"Dc Date","values":["2025"]},{"key":"dc:relation","label":"Dc Relation","values":["Electronic Theses & Dissertations"]},{"key":"dc:type","label":"Dc Type","values":["text","Electronic dissertations"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Artificial intelligence","Bioinformatics"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language","label":"Dc Language","values":["eng"]},{"key":"dc:rights","label":"Dc Rights","values":["Attribution 4.0 International"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier","label":"Identifier","values":["etd:52671","local:Ding_grad.msu_0128D_21640","https://doi.org/doi:10.25335/3r45-1546"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description","label":"Description","values":["Thesis (Ph.D.)--Michigan State University. Computer Science - Doctor of Philosophy, 2025","Single-cell transcriptomics enables high-resolution profiling of gene expression at the level of individual cells, offering unprecedented insights into cellular heterogeneity, development, and disease. However, the unique characteristics of single-cell data, such as high dimensionality, sparsity, and technical noise, pose significant computational challenges. Reproducibility of task-specific computational results is further hindered by limited open-sourcing and the lack of standardized, developer-friendly interfaces. Meanwhile, the rapid growth of public and private single-cell datasets demands methods that can scale effectively while preserving privacy. This thesis advances the field from task-specific tools to general-purpose foundation models through three key stages. First, I develop Graph Neural Networks based methods tailored to spatial transcriptomics. FOCUS introduces a semi-supervised graph contrastive learning framework that models the subcellular distribution of RNA transcripts to enhance spatial cell-type annotation while maintaining interpretability. GNNDECONVOLVER performs spot-level cell-type deconvolution using graph neural networks without relying on scRNA-seq reference data. Second, to enable systematic evaluation and broad usability, I build DANCE, a unified deep learning library and benchmarking platform for single-cell analysis. Extending this, DANCE 2.0 introduces automated and interpretable preprocessing that transforms the traditionally ad hoc trial-and-error approach into a transparent, data-driven workflow. Third, moving beyond task-specific solutions that require specific models and manual tuning for each task, I propose TABULA, a foundation model for single-cell genomics that enables self-supervised pretraining on large-scale and unlabeled datasets with built-in privacy safeguards. Unlike prior models that repurpose Natural Language Preprocessing techniques, TABULA is specifically designed for the structure and scale of single-cell data. Together, these contributions chart a cohesive path from task-specific models to a scalable, reproducible, and privacy-aware foundation model, laying the groundwork for unified and interpretable single-cell analysis.","Description based on online resource. Title from PDF t.p. (Michigan State University Fedora Repository, viewed ).","Includes bibliographical references."]},{"key":"dc:format","label":"Dc Format","values":["application/pdf","1 online resource (205 p.) : ill."]},{"key":"dc:title","label":"Title","values":["From Task-Specific Tools to Foundation Models in Single-Cell Genomics"]}]}],"canonical_facts":{"dc:contributor":["Tang, Jiliang","Qiu, Xiaojie","Xie, Yuying","Liu, Hui"],"dc:creator":["Ding, Jiayuan"],"dc:date":["2025"],"dc:description":["Thesis (Ph.D.)--Michigan State University. Computer Science - Doctor of Philosophy, 2025","Single-cell transcriptomics enables high-resolution profiling of gene expression at the level of individual cells, offering unprecedented insights into cellular heterogeneity, development, and disease. However, the unique characteristics of single-cell data, such as high dimensionality, sparsity, and technical noise, pose significant computational challenges. Reproducibility of task-specific computational results is further hindered by limited open-sourcing and the lack of standardized, developer-friendly interfaces. Meanwhile, the rapid growth of public and private single-cell datasets demands methods that can scale effectively while preserving privacy. This thesis advances the field from task-specific tools to general-purpose foundation models through three key stages. First, I develop Graph Neural Networks based methods tailored to spatial transcriptomics. FOCUS introduces a semi-supervised graph contrastive learning framework that models the subcellular distribution of RNA transcripts to enhance spatial cell-type annotation while maintaining interpretability. GNNDECONVOLVER performs spot-level cell-type deconvolution using graph neural networks without relying on scRNA-seq reference data. Second, to enable systematic evaluation and broad usability, I build DANCE, a unified deep learning library and benchmarking platform for single-cell analysis. Extending this, DANCE 2.0 introduces automated and interpretable preprocessing that transforms the traditionally ad hoc trial-and-error approach into a transparent, data-driven workflow. Third, moving beyond task-specific solutions that require specific models and manual tuning for each task, I propose TABULA, a foundation model for single-cell genomics that enables self-supervised pretraining on large-scale and unlabeled datasets with built-in privacy safeguards. Unlike prior models that repurpose Natural Language Preprocessing techniques, TABULA is specifically designed for the structure and scale of single-cell data. Together, these contributions chart a cohesive path from task-specific models to a scalable, reproducible, and privacy-aware foundation model, laying the groundwork for unified and interpretable single-cell analysis.","Description based on online resource. Title from PDF t.p. (Michigan State University Fedora Repository, viewed ).","Includes bibliographical references."],"dc:format":["application/pdf","1 online resource (205 p.) : ill."],"dc:identifier":["etd:52671","local:Ding_grad.msu_0128D_21640","https://doi.org/doi:10.25335/3r45-1546"],"dc:language":["eng"],"dc:relation":["Electronic Theses & Dissertations"],"dc:rights":["Attribution 4.0 International"],"dc:subject":["Artificial intelligence","Bioinformatics"],"dc:title":["From Task-Specific Tools to Foundation Models in Single-Cell Genomics"],"dc:type":["text","Electronic dissertations"]},"updated_at":"2026-11-18T18:59:26Z"}