Predicting the Interactions of Viral and Human Proteins

Simple item page
dc.contributor.authorEid, Fatma Elzahraa Sobhyen
dc.contributor.committeechairHeath, Lenwood S.en
dc.contributor.committeememberElHefnawi, Mahmoud M.en
dc.contributor.committeememberZhang, Liqingen
dc.contributor.committeememberOnufriev, Alexey V.en
dc.contributor.committeememberHuang, Berten
dc.contributor.departmentComputer Scienceen
dc.date.accessioned2017-05-04T08:00:34Zen
dc.date.available2017-05-04T08:00:34Zen
dc.date.issued2017-05-03en
dc.description.abstractThe world has proven unprepared for deadly viral outbreaks. Designing antiviral drugs and strategies requires a firm understanding of the interactions taken place between the proteins of the virus and human proteins. The current computational models for predicting these interactions consider only single viruses for which extensive prior knowledge is available. The two prediction frameworks in this dissertation, DeNovo and DeNovo-Human, make it possible for the first time to predict the interactions between any viral protein and human proteins. They further helped to answer critical questions about the Zika virus. DeNovo utilizes concepts from virology, bioinformatics, and machine learning to make predictions for novel viruses possible. It pools protein-protein interactions (PPIs) from different viruses sharing the same host. It further introduces taxonomic partitioning to make the reported performance reflect the situation of predicting for a novel virus. DeNovo avoids the expected low accuracy of such a prediction by introducing a negative sampling scheme that is based on sequence similarity. DeNovo achieved accuracy up to 81% and 86% when predicting for a new viral species and a new viral family, respectively. This result is comparable to the best achieved previously in single virus-host and intra-species PPI prediction cases. DeNovo predicts PPIs of a novel virus without requiring known PPIs for it, but with a limitation on the number of human proteins it can make predictions against. The second framework, DeNovo-Human, relaxes this limitation by forcing in-network prediction and random sampling while keeping the pooling technique of DeNovo. The accuracy and AUC are both promising ($>85%$, and $>91%$ respectively). DeNovo-Human facilitates predicting the virus-human PPI network. To demonstrate how the two frameworks can enrich our knowledge about virus behavior, I use them to answer interesting questions about the Zika virus. The research questions examine how the Zika virus enters human cells, fights the innate immune system, and causes microcephaly. The answers obtained are well supported by recently published Zika virus studies.en
dc.description.abstractgeneralWhen a virus attacks a human body, it disturbs the host cells by interacting with their proteins. Identifying these interactions is key to fighting the virus. In this dissertation, I developed two computational tools to identify the interactions for any virus infecting the human. I further used these tools to answer interesting questions about the Zika virus behavior. The results are in agreement with recently published experimental studies about the virus.en
dc.description.degreePh. D.en
dc.format.mediumETDen
dc.identifier.othervt_gsexam:11072en
dc.identifier.urihttp://hdl.handle.net/10919/77581en
dc.publisherVirginia Techen
dc.rightsIn Copyrighten
dc.rights.urihttp://rightsstatements.org/vocab/InC/1.0/en
dc.subjectProtein-Protein Interactionen
dc.subjectVirusen
dc.subjectMachine learningen
dc.subjectZika Virusen
dc.titlePredicting the Interactions of Viral and Human Proteinsen
dc.typeDissertationen
thesis.degree.disciplineComputer Science and Applicationsen
thesis.degree.grantorVirginia Polytechnic Institute and State Universityen
thesis.degree.leveldoctoralen
thesis.degree.namePh. D.en

Files

Original bundle
Now showing 1 - 1 of 1
Thumbnail Image
Name:
Eid_FS_D_2017.pdf
Size:
7.46 MB
Format:
Adobe Portable Document Format
Download

Collections

Doctoral Dissertations