Identifying the causal interactions between genes and their proteins during the differentiation of specialized cells such as mechanosensory hair cells in vertebrates' inner ears and fishes' lateral lines requires an accurate description of the time-lagged relationships between transcription factors and their target genes. Here I describe Depicting Lagged Causality (DELAY), a convolutional neural network for the inference of gene-regulatory relationships across pseudotime-ordered single-cell trajectories. I first show that combining supervised deep learning with joint probability matrices of pseudotime-lagged trajectories allows the neural network to overcome important limitations of ordinary Granger causality-based methods, for example, the inability to infer cyclic relationships such as feedback loops. The algorithm outperforms several common methods for inferring gene regulation and, when given partial ground-truth labels, predicts novel gene-regulatory networks from single-cell RNA sequencing and single-cell ATAC sequencing data sets. To validate this approach, I use DELAY to identify important genes and modules in the regulatory network for auditory hair cell development in the murine inner ear, as well as likely DNA-binding partners for two hair cell cofactors (Hist1h1c and Ccnd1) and a novel DNA-binding sequence for the transcription factor Fiz1. In zebrafish, lateral-line neuromasts can regenerate damaged hair cells by expressing genes such as atoh1a—the master regulator of hair cell fate—in progenitors known as supporting cells. To identify adaptations that promote the rapid regeneration of hair cells in larval zebrafish, I also use DELAY to infer regenerating neuromasts' early gene-regulatory network. The central hub in the network, Y-box binding protein 1 (ybx1), is highly expressed in hair cell progenitors and young hair cells and its protein can recognize binding sites in the candidate regeneration-responsive promoter element for atoh1a. I show that neuromasts from ybx1 mutant zebrafish larvae display consistent, regeneration-specific deficits in hair cell number and initiate both hair cell regeneration and atoh1a expression 20% slower than in siblings. By demonstrating that ybx1 promotes rapid hair cell regeneration in neuromasts through early atoh1a upregulation, these results strongly support DELAY's ability to identify key regulators of gene expression dynamics. I provide a user-friendly implementation of DELAY under an open-source license at https://github.com/calebclayreagor/DELAY.
","abstract_html":"<p>Identifying the causal interactions between genes and their proteins during the differentiation of specialized cells such as mechanosensory hair cells in vertebrates' inner ears and fishes' lateral lines requires an accurate description of the time-lagged relationships between transcription factors and their target genes. Here I describe <em>De</em>picting <em>Lagged</em> Causality (DELAY), a convolutional neural network for the inference of gene-regulatory relationships across pseudotime-ordered single-cell trajectories. I first show that combining supervised deep learning with joint probability matrices of pseudotime-lagged trajectories allows the neural network to overcome important limitations of ordinary Granger causality-based methods, for example, the inability to infer cyclic relationships such as feedback loops. The algorithm outperforms several common methods for inferring gene regulation and, when given partial ground-truth labels, predicts novel gene-regulatory networks from single-cell RNA sequencing and single-cell ATAC sequencing data sets. To validate this approach, I use DELAY to identify important genes and modules in the regulatory network for auditory hair cell development in the murine inner ear, as well as likely DNA-binding partners for two hair cell cofactors (Hist1h1c and Ccnd1) and a novel DNA-binding sequence for the transcription factor Fiz1. In zebrafish, lateral-line neuromasts can regenerate damaged hair cells by expressing genes such as <em>atoh1a</em>—the master regulator of hair cell fate—in progenitors known as supporting cells. To identify adaptations that promote the rapid regeneration of hair cells in larval zebrafish, I also use DELAY to infer regenerating neuromasts' early gene-regulatory network. The central hub in the network, <em>Y-box binding protein 1 (ybx1)</em>, is highly expressed in hair cell progenitors and young hair cells and its protein can recognize binding sites in the candidate regeneration-responsive promoter element for <em>atoh1a</em>. I show that neuromasts from <em>ybx1 </em>mutant zebrafish larvae display consistent, regeneration-specific deficits in hair cell number and initiate both hair cell regeneration and <em>atoh1a</em> expression 20% slower than in siblings. By demonstrating that <em>ybx1</em> promotes rapid hair cell regeneration in neuromasts through early <em>atoh1a</em> upregulation, these results strongly support DELAY's ability to identify key regulators of gene expression dynamics. I provide a user-friendly implementation of DELAY under an open-source license at https://github.com/calebclayreagor/DELAY.</p>","abstract_has_math":false,"creators":["Reagor, Caleb C"],"institution":null,"degree_name":"Doctor of Philosophy (PhD)","degree_level":"Thesis","degree_discipline":null,"degree_department":null,"school":null,"contributors":["A. James Hudspeth"],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2025,"date_issued":"2025-01-01T08:00:00Z","date_published":"2025-01-01T08:00:00Z","updated_at":"2026-07-24T04:11:51Z","subjects":["gene regulation","pseudotime","deep learning","single-cell RNA sequencing (scRNA-seq)","hair cell differentiation","regeneration","Life Sciences"],"languages":[],"rights":[],"rights_urls":[],"identifier_entries":[]},"links":{"outbound_url":"https://digitalcommons.rockefeller.edu/student_theses_and_dissertations/824","outbound_label":"Repository record","outbound_source":"dc:identifier"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor","label":"Contributor","values":["A. James Hudspeth"]},{"key":"dc:creator","label":"Author","values":["Reagor, Caleb C"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"thesis:degree_level","label":"Degree Level","values":["Thesis"]},{"key":"thesis:degree_name","label":"Degree Name","values":["Doctor of Philosophy (PhD)"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["gene regulation","pseudotime","deep learning","single-cell RNA sequencing (scRNA-seq)","hair cell differentiation","regeneration","Life Sciences"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier","label":"Identifier","values":["https://digitalcommons.rockefeller.edu/student_theses_and_dissertations/824"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["Identifying the causal interactions between genes and their proteins during the differentiation of specialized cells such as mechanosensory hair cells in vertebrates' inner ears and fishes' lateral lines requires an accurate description of the time-lagged relationships between transcription factors and their target genes. Here I describe Depicting Lagged Causality (DELAY), a convolutional neural network for the inference of gene-regulatory relationships across pseudotime-ordered single-cell trajectories. I first show that combining supervised deep learning with joint probability matrices of pseudotime-lagged trajectories allows the neural network to overcome important limitations of ordinary Granger causality-based methods, for example, the inability to infer cyclic relationships such as feedback loops. The algorithm outperforms several common methods for inferring gene regulation and, when given partial ground-truth labels, predicts novel gene-regulatory networks from single-cell RNA sequencing and single-cell ATAC sequencing data sets. To validate this approach, I use DELAY to identify important genes and modules in the regulatory network for auditory hair cell development in the murine inner ear, as well as likely DNA-binding partners for two hair cell cofactors (Hist1h1c and Ccnd1) and a novel DNA-binding sequence for the transcription factor Fiz1. In zebrafish, lateral-line neuromasts can regenerate damaged hair cells by expressing genes such as atoh1a—the master regulator of hair cell fate—in progenitors known as supporting cells. To identify adaptations that promote the rapid regeneration of hair cells in larval zebrafish, I also use DELAY to infer regenerating neuromasts' early gene-regulatory network. The central hub in the network, Y-box binding protein 1 (ybx1), is highly expressed in hair cell progenitors and young hair cells and its protein can recognize binding sites in the candidate regeneration-responsive promoter element for atoh1a. I show that neuromasts from ybx1 mutant zebrafish larvae display consistent, regeneration-specific deficits in hair cell number and initiate both hair cell regeneration and atoh1a expression 20% slower than in siblings. By demonstrating that ybx1 promotes rapid hair cell regeneration in neuromasts through early atoh1a upregulation, these results strongly support DELAY's ability to identify key regulators of gene expression dynamics. I provide a user-friendly implementation of DELAY under an open-source license at https://github.com/calebclayreagor/DELAY.
"]},{"key":"dc:title","label":"Title","values":["Decoding Dynamic Gene Regulation in Hair Cell Development and Regeneration"]}]}],"canonical_facts":{"dc:contributor":["A. James Hudspeth"],"dc:creator":["Reagor, Caleb C"],"dc:description.abstract":["Identifying the causal interactions between genes and their proteins during the differentiation of specialized cells such as mechanosensory hair cells in vertebrates' inner ears and fishes' lateral lines requires an accurate description of the time-lagged relationships between transcription factors and their target genes. Here I describe Depicting Lagged Causality (DELAY), a convolutional neural network for the inference of gene-regulatory relationships across pseudotime-ordered single-cell trajectories. I first show that combining supervised deep learning with joint probability matrices of pseudotime-lagged trajectories allows the neural network to overcome important limitations of ordinary Granger causality-based methods, for example, the inability to infer cyclic relationships such as feedback loops. The algorithm outperforms several common methods for inferring gene regulation and, when given partial ground-truth labels, predicts novel gene-regulatory networks from single-cell RNA sequencing and single-cell ATAC sequencing data sets. To validate this approach, I use DELAY to identify important genes and modules in the regulatory network for auditory hair cell development in the murine inner ear, as well as likely DNA-binding partners for two hair cell cofactors (Hist1h1c and Ccnd1) and a novel DNA-binding sequence for the transcription factor Fiz1. In zebrafish, lateral-line neuromasts can regenerate damaged hair cells by expressing genes such as atoh1a—the master regulator of hair cell fate—in progenitors known as supporting cells. To identify adaptations that promote the rapid regeneration of hair cells in larval zebrafish, I also use DELAY to infer regenerating neuromasts' early gene-regulatory network. The central hub in the network, Y-box binding protein 1 (ybx1), is highly expressed in hair cell progenitors and young hair cells and its protein can recognize binding sites in the candidate regeneration-responsive promoter element for atoh1a. I show that neuromasts from ybx1 mutant zebrafish larvae display consistent, regeneration-specific deficits in hair cell number and initiate both hair cell regeneration and atoh1a expression 20% slower than in siblings. By demonstrating that ybx1 promotes rapid hair cell regeneration in neuromasts through early atoh1a upregulation, these results strongly support DELAY's ability to identify key regulators of gene expression dynamics. I provide a user-friendly implementation of DELAY under an open-source license at https://github.com/calebclayreagor/DELAY.
"],"dc:identifier":["https://digitalcommons.rockefeller.edu/student_theses_and_dissertations/824"],"dc:subject":["gene regulation","pseudotime","deep learning","single-cell RNA sequencing (scRNA-seq)","hair cell differentiation","regeneration","Life Sciences"],"dc:title":["Decoding Dynamic Gene Regulation in Hair Cell Development and Regeneration"],"thesis:degree_level":["Thesis"],"thesis:degree_name":["Doctor of Philosophy (PhD)"]},"updated_at":"2026-07-24T04:11:51Z"}