Predicting Immunogenic CD4+ T Cell Epitopes in Bacteria Using Antigen and Peptide Features DOI Creative Commons
Daniel Marrama, Hannah Battey, Ehdieh Khaledian

et al.

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: Oct. 29, 2024

Abstract Background T cell epitope prediction methods have been broadly utilized to facilitate discovery in infectious agents and help design reagents, diagnostics, vaccines. Current are mainly focused on peptide presentation by MHC molecules, which is a necessary but not sufficient requirement for an epitope. For complex pathogens such as bacteria, it would be desirable make predictions more specific limit the number of candidates that experimentally tested. Objective To develop machine learning-based model integrates both peptide-level antigen-level features improve specificity CD4+ bacteria. Methods We used dataset 20,216 peptides from Mycobacterium tuberculosis (Mtb), tested recognition Mtb-infected participants, led n = 144 epitopes. each peptide, we calculated six (e.g. class II binding conservation scores) including RNA expression levels subcellular localization scores). Three learning algorithms—Random Forest, Gradient Boosting, XGBoost—were trained using stratified, 5-fold cross-validation combined into ensemble model. Experimental validation was performed Streptococcus pneumoniae peptides, ex vivo IFNγ assays confirm predictive performance. Results The achieved ROC-AUC 0.91 predicting immunogenic (Mtb) dataset. Gene were identified most impactful features, followed predictions. When validated independent Bordetella pertussis dataset, demonstrated accurate capability, especially with broad participant cohort (ROC-AUC up 0.82). Prospectively applying , synthesized predicted our or non-immunogenic. Ex testing PBMCs healthy participants showed elicited significantly higher responses than non-immunogenic validating Conclusions Our approach, integrating antigen effectively predicts epitopes across different bacterial pathogens. This method enhances selection efficiency, aiding vaccine development immunological research reducing need extensive experimental screening.

Language: Английский

Predicting Immunogenic CD4+ T Cell Epitopes in Bacteria Using Antigen and Peptide Features DOI Creative Commons
Daniel Marrama, Hannah Battey, Ehdieh Khaledian

et al.

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: Oct. 29, 2024

Abstract Background T cell epitope prediction methods have been broadly utilized to facilitate discovery in infectious agents and help design reagents, diagnostics, vaccines. Current are mainly focused on peptide presentation by MHC molecules, which is a necessary but not sufficient requirement for an epitope. For complex pathogens such as bacteria, it would be desirable make predictions more specific limit the number of candidates that experimentally tested. Objective To develop machine learning-based model integrates both peptide-level antigen-level features improve specificity CD4+ bacteria. Methods We used dataset 20,216 peptides from Mycobacterium tuberculosis (Mtb), tested recognition Mtb-infected participants, led n = 144 epitopes. each peptide, we calculated six (e.g. class II binding conservation scores) including RNA expression levels subcellular localization scores). Three learning algorithms—Random Forest, Gradient Boosting, XGBoost—were trained using stratified, 5-fold cross-validation combined into ensemble model. Experimental validation was performed Streptococcus pneumoniae peptides, ex vivo IFNγ assays confirm predictive performance. Results The achieved ROC-AUC 0.91 predicting immunogenic (Mtb) dataset. Gene were identified most impactful features, followed predictions. When validated independent Bordetella pertussis dataset, demonstrated accurate capability, especially with broad participant cohort (ROC-AUC up 0.82). Prospectively applying , synthesized predicted our or non-immunogenic. Ex testing PBMCs healthy participants showed elicited significantly higher responses than non-immunogenic validating Conclusions Our approach, integrating antigen effectively predicts epitopes across different bacterial pathogens. This method enhances selection efficiency, aiding vaccine development immunological research reducing need extensive experimental screening.

Language: Английский

Citations

0