{"id":{"repo_id":"uic","oai_identifier":"oai:figshare.com:article/32994986"},"canonical_url":"https://search.dev.ndltd.org/etd/uic/oai:figshare.com:article/32994986","repository":{"repo_id":"uic","name":"University of Illinois - Chicago","base_url":"https://api.figshare.com/v2/oai"},"display":{"title":"Knowledge-Guided Machine Learning for Single-Cell Regulatory Genomics","abstract":"Transcription factors (TFs) and cis-regulatory elements coordinate gene regulation, and single-cell sequencing now enables these programs to be studied at high resolution. However, single-cell RNA-seq and ATAC-seq data are sparse and high-dimensional, making it difficult for existing methods to reliably infer TF activity and gene-regulatory networks. This defense presents a set of knowledge-guided machine-learning approaches that embed prior biological evidence into modern analytical models to improve regulatory inference from noisy single-cell data. First, we evaluate TF-IDF transformations and dimensionality-reduction methods for information retrieval in scATAC-seq, showing TF-IDF consistently improves clustering and feature extraction and performs best when paired with variational autoencoders. Second, we extend the BITFAM framework to jointly model scRNA-seq and scATAC-seq in a skin wound-healing study, identifying early and late macrophage subpopulations, cooperative TF communities, and supporting a pro-inflammatory role for NR4A1 with in vivo validation. Third, we introduce scRegulate, a VAE that embeds TF-target priors to infer TF activities and context-specific regulatory networks from scRNA-seq, with benchmarking demonstrating improved recovery of perturbation effects and cell-type-specific programs while scaling approximately linearly with dataset size. Finally, we present RAGulate, a retrieval-augmented generation system that links predicted TF-target relationships to supporting literature and produces an evidence-based confidence score and explanation. Together, these contributions advance interpretable, prior-informed modeling for single-cell regulatory genomics.","abstract_html":"Transcription factors (TFs) and cis-regulatory elements coordinate gene regulation, and single-cell sequencing now enables these programs to be studied at high resolution. However, single-cell RNA-seq and ATAC-seq data are sparse and high-dimensional, making it difficult for existing methods to reliably infer TF activity and gene-regulatory networks. This defense presents a set of knowledge-guided machine-learning approaches that embed prior biological evidence into modern analytical models to improve regulatory inference from noisy single-cell data. First, we evaluate TF-IDF transformations and dimensionality-reduction methods for information retrieval in scATAC-seq, showing TF-IDF consistently improves clustering and feature extraction and performs best when paired with variational autoencoders. Second, we extend the BITFAM framework to jointly model scRNA-seq and scATAC-seq in a skin wound-healing study, identifying early and late macrophage subpopulations, cooperative TF communities, and supporting a pro-inflammatory role for NR4A1 with in vivo validation. Third, we introduce scRegulate, a VAE that embeds TF-target priors to infer TF activities and context-specific regulatory networks from scRNA-seq, with benchmarking demonstrating improved recovery of perturbation effects and cell-type-specific programs while scaling approximately linearly with dataset size. Finally, we present RAGulate, a retrieval-augmented generation system that links predicted TF-target relationships to supporting literature and produces an evidence-based confidence score and explanation. Together, these contributions advance interpretable, prior-informed modeling for single-cell regulatory genomics.","abstract_has_math":false,"creators":["Mehrdad Zandigohar (24399968)"],"institution":null,"degree_name":null,"degree_level":null,"degree_discipline":null,"degree_department":null,"school":null,"contributors":[],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2026,"date_issued":"2026-05-01T00:00:00Z","date_published":"2026-05-01T00:00:00Z","updated_at":"2026-07-27T21:33:46Z","subjects":["Bioinformatics","Computational biology","Genomics"],"languages":[],"rights":["In Copyright","Open Access after 2028-05-01"],"rights_urls":[],"identifier_entries":[]},"links":{"outbound_url":"https://doi.org/10.25417/uic.32994986.v1","outbound_label":"DOI","outbound_source":"dc:identifier"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:creator","label":"Author","values":["Mehrdad Zandigohar (24399968)"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date","label":"Dc Date","values":["2026-05-01T00:00:00Z"]},{"key":"dc:relation","label":"Dc Relation","values":["https://figshare.com/articles/thesis/Knowledge-Guided_Machine_Learning_for_Single-Cell_Regulatory_Genomics/32994986"]},{"key":"dc:type","label":"Dc Type","values":["Text","Thesis"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Bioinformatics","Computational biology","Genomics"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:rights","label":"Dc Rights","values":["In Copyright","Open Access after 2028-05-01"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier","label":"Identifier","values":["10.25417/uic.32994986.v1"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description","label":"Description","values":["Transcription factors (TFs) and cis-regulatory elements coordinate gene regulation, and single-cell sequencing now enables these programs to be studied at high resolution. However, single-cell RNA-seq and ATAC-seq data are sparse and high-dimensional, making it difficult for existing methods to reliably infer TF activity and gene-regulatory networks. This defense presents a set of knowledge-guided machine-learning approaches that embed prior biological evidence into modern analytical models to improve regulatory inference from noisy single-cell data. First, we evaluate TF-IDF transformations and dimensionality-reduction methods for information retrieval in scATAC-seq, showing TF-IDF consistently improves clustering and feature extraction and performs best when paired with variational autoencoders. Second, we extend the BITFAM framework to jointly model scRNA-seq and scATAC-seq in a skin wound-healing study, identifying early and late macrophage subpopulations, cooperative TF communities, and supporting a pro-inflammatory role for NR4A1 with in vivo validation. Third, we introduce scRegulate, a VAE that embeds TF-target priors to infer TF activities and context-specific regulatory networks from scRNA-seq, with benchmarking demonstrating improved recovery of perturbation effects and cell-type-specific programs while scaling approximately linearly with dataset size. Finally, we present RAGulate, a retrieval-augmented generation system that links predicted TF-target relationships to supporting literature and produces an evidence-based confidence score and explanation. Together, these contributions advance interpretable, prior-informed modeling for single-cell regulatory genomics."]},{"key":"dc:title","label":"Title","values":["Knowledge-Guided Machine Learning for Single-Cell Regulatory Genomics"]}]}],"canonical_facts":{"dc:creator":["Mehrdad Zandigohar (24399968)"],"dc:date":["2026-05-01T00:00:00Z"],"dc:description":["Transcription factors (TFs) and cis-regulatory elements coordinate gene regulation, and single-cell sequencing now enables these programs to be studied at high resolution. However, single-cell RNA-seq and ATAC-seq data are sparse and high-dimensional, making it difficult for existing methods to reliably infer TF activity and gene-regulatory networks. This defense presents a set of knowledge-guided machine-learning approaches that embed prior biological evidence into modern analytical models to improve regulatory inference from noisy single-cell data. First, we evaluate TF-IDF transformations and dimensionality-reduction methods for information retrieval in scATAC-seq, showing TF-IDF consistently improves clustering and feature extraction and performs best when paired with variational autoencoders. Second, we extend the BITFAM framework to jointly model scRNA-seq and scATAC-seq in a skin wound-healing study, identifying early and late macrophage subpopulations, cooperative TF communities, and supporting a pro-inflammatory role for NR4A1 with in vivo validation. Third, we introduce scRegulate, a VAE that embeds TF-target priors to infer TF activities and context-specific regulatory networks from scRNA-seq, with benchmarking demonstrating improved recovery of perturbation effects and cell-type-specific programs while scaling approximately linearly with dataset size. Finally, we present RAGulate, a retrieval-augmented generation system that links predicted TF-target relationships to supporting literature and produces an evidence-based confidence score and explanation. Together, these contributions advance interpretable, prior-informed modeling for single-cell regulatory genomics."],"dc:identifier":["10.25417/uic.32994986.v1"],"dc:relation":["https://figshare.com/articles/thesis/Knowledge-Guided_Machine_Learning_for_Single-Cell_Regulatory_Genomics/32994986"],"dc:rights":["In Copyright","Open Access after 2028-05-01"],"dc:subject":["Bioinformatics","Computational biology","Genomics"],"dc:title":["Knowledge-Guided Machine Learning for Single-Cell Regulatory Genomics"],"dc:type":["Text","Thesis"]},"updated_at":"2026-07-27T21:33:46Z"}