Bacteriophages are the most ubiquitous biological entity on the planet, but most viruses found in nature cannot be cultured in the laboratory and encode genes whose sequences lack similarity with current nucleotide and protein databases. New predictive methods are thus necessary to determine the phenotype of viruses. In this work, we leverage the physical geometrical constraints of viruses to quantify the correlation between the geometric and genomic characteristics of tailed phages, and predict physical features such as architecture and genome length of uncultured viruses using allometric models and machine learning algorithms. Here, we present a model to predict the T-number as a measure of capsid architecture of tailed phages based on the genome length of the virus, and another to predict viral genome length from their structural genes such as the Major Capsid Protein (MCP). Our research indicates that the genome length can predict capsid architecture with 90% accuracy, and that MCP features can predict capsid architecture with an overall 84% accuracy. Using the same MCP features in a multi-step predictive model predict the genome length of a virus with an overall average mean relative error of 7.6%. Since this model is based on a single gene, improvement may be achieved with the addition of other structural genes, such as the portal or scaffolding genes, or by refining our methods of isolating the MCP. This approach can help predict the phenotype of uncultured viruses and fill the gap in our understanding of the virosphere.
","abstract_html":"<p>Bacteriophages are the most ubiquitous biological entity on the planet, but most viruses found in nature cannot be cultured in the laboratory and encode genes whose sequences lack similarity with current nucleotide and protein databases. New predictive methods are thus necessary to determine the phenotype of viruses. In this work, we leverage the physical geometrical constraints of viruses to quantify the correlation between the geometric and genomic characteristics of tailed phages, and predict physical features such as architecture and genome length of uncultured viruses using allometric models and machine learning algorithms. Here, we present a model to predict the T-number as a measure of capsid architecture of tailed phages based on the genome length of the virus, and another to predict viral genome length from their structural genes such as the Major Capsid Protein (MCP). Our research indicates that the genome length can predict capsid architecture with 90% accuracy, and that MCP features can predict capsid architecture with an overall 84% accuracy. Using the same MCP features in a multi-step predictive model predict the genome length of a virus with an overall average mean relative error of 7.6%. Since this model is based on a single gene, improvement may be achieved with the addition of other structural genes, such as the portal or scaffolding genes, or by refining our methods of isolating the MCP. This approach can help predict the phenotype of uncultured viruses and fill the gap in our understanding of the virosphere.</p>","abstract_has_math":false,"creators":["Lee, Diana Yvette"],"institution":null,"degree_name":"Computational Science Joint PhD with San Diego State University, PhD","degree_level":"Open Access Dissertation","degree_discipline":"Institute of Mathematical Sciences","degree_department":null,"school":null,"contributors":["Anca Segall","Manal Swairjo","Allon Percus & Marina Chugunova"],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2024,"date_issued":"2024-01-01T08:00:00Z","date_published":"2024-01-01T08:00:00Z","updated_at":"2026-07-24T01:41:01Z","subjects":["bacteriophages","computational biology","machine learning","structural biology","viral architecture","viral biomathematics","Biology","Computer Sciences"],"languages":[],"rights":[],"rights_urls":[],"identifier_entries":[]},"links":{"outbound_url":"https://scholarship.claremont.edu/cgu_etd/906","outbound_label":"Repository record","outbound_source":"dc:identifier"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor","label":"Contributor","values":["Anca Segall","Manal Swairjo","Allon Percus & Marina Chugunova"]},{"key":"dc:creator","label":"Author","values":["Lee, Diana Yvette"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.available","label":"Dc Date Available","values":["2025-01-22T08:00:00Z"]},{"key":"thesis:degree_discipline","label":"Discipline","values":["Institute of Mathematical Sciences"]},{"key":"thesis:degree_level","label":"Degree Level","values":["Open Access Dissertation"]},{"key":"thesis:degree_name","label":"Degree Name","values":["Computational Science Joint PhD with San Diego State University, PhD"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["bacteriophages","computational biology","machine learning","structural biology","viral architecture","viral biomathematics","Biology","Computer Sciences"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier","label":"Identifier","values":["https://scholarship.claremont.edu/cgu_etd/906"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["Bacteriophages are the most ubiquitous biological entity on the planet, but most viruses found in nature cannot be cultured in the laboratory and encode genes whose sequences lack similarity with current nucleotide and protein databases. New predictive methods are thus necessary to determine the phenotype of viruses. In this work, we leverage the physical geometrical constraints of viruses to quantify the correlation between the geometric and genomic characteristics of tailed phages, and predict physical features such as architecture and genome length of uncultured viruses using allometric models and machine learning algorithms. Here, we present a model to predict the T-number as a measure of capsid architecture of tailed phages based on the genome length of the virus, and another to predict viral genome length from their structural genes such as the Major Capsid Protein (MCP). Our research indicates that the genome length can predict capsid architecture with 90% accuracy, and that MCP features can predict capsid architecture with an overall 84% accuracy. Using the same MCP features in a multi-step predictive model predict the genome length of a virus with an overall average mean relative error of 7.6%. Since this model is based on a single gene, improvement may be achieved with the addition of other structural genes, such as the portal or scaffolding genes, or by refining our methods of isolating the MCP. This approach can help predict the phenotype of uncultured viruses and fill the gap in our understanding of the virosphere.
"]},{"key":"dc:title","label":"Title","values":["Reaching Across the Divide: Tools for Bridging Structural and Viral Genomics Using a Combination of Biophysical Principles and Machine Learning"]}]}],"canonical_facts":{"dc:contributor":["Anca Segall","Manal Swairjo","Allon Percus & Marina Chugunova"],"dc:creator":["Lee, Diana Yvette"],"dc:date.available":["2025-01-22T08:00:00Z"],"dc:description.abstract":["Bacteriophages are the most ubiquitous biological entity on the planet, but most viruses found in nature cannot be cultured in the laboratory and encode genes whose sequences lack similarity with current nucleotide and protein databases. New predictive methods are thus necessary to determine the phenotype of viruses. In this work, we leverage the physical geometrical constraints of viruses to quantify the correlation between the geometric and genomic characteristics of tailed phages, and predict physical features such as architecture and genome length of uncultured viruses using allometric models and machine learning algorithms. Here, we present a model to predict the T-number as a measure of capsid architecture of tailed phages based on the genome length of the virus, and another to predict viral genome length from their structural genes such as the Major Capsid Protein (MCP). Our research indicates that the genome length can predict capsid architecture with 90% accuracy, and that MCP features can predict capsid architecture with an overall 84% accuracy. Using the same MCP features in a multi-step predictive model predict the genome length of a virus with an overall average mean relative error of 7.6%. Since this model is based on a single gene, improvement may be achieved with the addition of other structural genes, such as the portal or scaffolding genes, or by refining our methods of isolating the MCP. This approach can help predict the phenotype of uncultured viruses and fill the gap in our understanding of the virosphere.
"],"dc:identifier":["https://scholarship.claremont.edu/cgu_etd/906"],"dc:subject":["bacteriophages","computational biology","machine learning","structural biology","viral architecture","viral biomathematics","Biology","Computer Sciences"],"dc:title":["Reaching Across the Divide: Tools for Bridging Structural and Viral Genomics Using a Combination of Biophysical Principles and Machine Learning"],"thesis:degree_discipline":["Institute of Mathematical Sciences"],"thesis:degree_level":["Open Access Dissertation"],"thesis:degree_name":["Computational Science Joint PhD with San Diego State University, PhD"]},"updated_at":"2026-07-24T01:41:01Z"}