Skip to content

Computational Models of Immunity to Pertussis Booster vaccinations (CMI-PB) Project to generate experimental data to create and test computational models that predict vaccination outcomes based on the baseline state of the vaccines.

Notifications You must be signed in to change notification settings

joe-jhou2/CMI-PB-Challenge

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

8 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

CMI-PB Challenge Summary

Author: Joe Hou (Fred Hutch), Yunda Huang(Fred Hutch), James Kobie (University of Alabama)

The CMI-PB Challenge is centered around the analysis of immune responses to Pertussis booster vaccinations using systems vaccinology. Participants are provided with multi-omics datasets from a cohort of individuals primed in infancy with either acellular Pertussis (aP) or whole cell Pertussis (wP) vaccines and later boosted with Tdap. Individuals born before 1995 received wP, while those after 1996 received aP. The study design includes pre- and post-booster blood and plasma samples at intervals of 1, 3, 7, 14 days and more later time points.

The datasets encompass:

(1) Cell frequency in PBMCs analyzed by flow cytometry (2) Gene expression profiles covering over 50,000 genes (3) Plasma cytokine concentrations for 30 soluble proteins measured with Olink technology (4) Antibody titers against more than 7 antigens

Our Approach:

In addressing the complexities of the CMI-PB Challenge, we opted for the SuperLearner algorithm. SuperLearner is an ensemble machine learning method that combines multiple prediction models to improve the accuracy of predictions. This approach is particularly advantageous in our context for several reasons:

Advantages of SuperLearner:

  • Ensemble Approach: SuperLearner doesn't rely on a single model but instead combines several models. This ensemble approach often results in better prediction accuracy, as it leverages the strengths of various algorithms.

  • Flexibility: It allows for the integration of diverse types of models, including both parametric and non-parametric approaches. This flexibility is crucial in handling the multifaceted nature of our multi-omics datasets.

  • Customizability: SuperLearner can be tailored to our specific problem. We can choose which algorithms to include in the ensemble, allowing us to fine-tune the model to our data's unique characteristics.

  • Performance Evaluation: It offers a robust cross-validation framework, enabling us to assess the performance of individual models and the ensemble as a whole. This feature helps in identifying the most effective combination of models for our prediction task.

  • Handling Complex Interactions: Given the complexity of immunological data, SuperLearner is well-suited to capture intricate patterns and interactions within the data, which might be missed by simpler models.

By leveraging SuperLearner, we aim to harness these benefits to effectively analyze the immune response data and draw meaningful conclusions that could contribute to advancements in vaccine research.

Key steps in my methodology include:

  • Data Description and Preprocessing: Thoroughly examining the baseline characteristics of the multi-omics datasets to ensure a robust starting point for analysis.

  • Consistency Check: Ensuring data integrity and consistency across different omics datasets, which is crucial for reliable model building.

  • Model Building with SuperLearner:

    -Selecting multiple prediction methods within SuperLearner to handle the complex nature of the immune response data. -Carefully choosing relevant data features, including considering predictions based on the target assay or incorporating additional variables from other assays. -Finalizing the model with optimized selection of methods and data inputs to accurately predict the outcomes of the Tdap booster vaccination.

SuperLearner Approach

This comprehensive approach, rooted in the capabilities of SuperLearner, is designed to untangle the intricate patterns in the immune response data, aiming to contribute meaningful insights into vaccine-induced immunity.

About

Computational Models of Immunity to Pertussis Booster vaccinations (CMI-PB) Project to generate experimental data to create and test computational models that predict vaccination outcomes based on the baseline state of the vaccines.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published